Embed Size (px)
Citation preview
I
The Determinants and Evolution of Major Inter-firm
Transactions in the U.S. Apparel Sector
Xiao (Mimosa) Zhao
Supervisor: Margaret Dalziel
Thesis submitted to the
Faculty of Graduate and Postdoctoral Studies
in partial fulfillment of the requirements
for the M.Sc. degree in Management
Telfer School of Management
University of Ottawa
© Xiao Zhao, Ottawa, Canada, 2013
II
Acknowledgements
I express my gratitude to those who help me to complete my thesis and my Master degree
in Telfer School of Management. I will be forever grateful for the tremendous supervision,
support, and encouragement that they collectively and individually provided to me
throughout my journey at Telfer.
First, I would like to sincerely thank my beautiful thesis supervisor Margaret Dalziel.
Working with her has become one of the most splendid experiences in my life. Margaret
did not only help me with research and thesis, but also provided enormous advice on my
life and career, which helped me through many challenges, and which I appreciate deeply.
I enjoyed the chatting with her on topics about choice of life and career, jazz music, China,
travels, and even good food. To me, Margaret is much more than an advisor who
supervised my thesis, but also a great friend and a family. She is a mentor for my research
as well as a role model for my life.
I am also grateful to my insightful and generous committee members – Prof. Freel and Prof.
Chamberlin. Thanks for the valuable comments from them on my thesis. I am also indebted
to Prof. Ajax who supported me to ASAC conference and helped on my Statistics course.
I would like to thank all my friends in Canada who make my life much easier and happier
and share the time with me. Especially, the encouragement and support from Wangzhe Li
and Yuxuan Song have helped me to go through many challenges. I would also like to
thank Prof. Yang and Prof. Luo who helped me with my thesis in Waterloo.
My gratitude also goes to my cousin's family. Thanks to their support in Toronto and
provide me with a second home. I enjoyed the Christmas holiday with them.
My thesis is dedicated to my parents. Without their love and support, I will never achieve it
III
and enjoy the happiness of life when I am alone far away from home.
感谢我的爸爸妈妈，教会我享受生活，开心做人。
IV
Abstract
This study provides a systematic description of the nature and evolution of major
transactions in the U.S. apparel sector, using a theory that applies across sectors. This
research investigates the determinants of the existence and magnitude of major inter-firm
transactions, relying on a unique longitudinal dataset of over 2,000 of the largest
transactional (buy-sell) relations between publicly traded firms in the U.S. apparel sector.
The results indicate the importance of inter-firm complementarity, rather than inter-firm
similarity, in explaining the sector architecture; thus contributing to the future
improvement of industry classification systems. This study also contributes to a deeper
understanding of the apparel sector focusing on the change in the relative importance of
manufacturing activities versus service activities and in the involvement of firms from the
outside apparel sector. Implications of inter-firm transactions are discussed regarding
industry policies, and human and environmental welfare in manufacturing and raw
materials industries.
V
Table of Contents
Acknowledgements ...................................................................................................... II
Abstract ...................................................................................................................... IV
Table of Contents ......................................................................................................... V
List of Tables ............................................................................................................. VII
1. Introduction ...........................................................................................................1
2. Literature Review ..................................................................................................7
2.1 Theories that Explain Firm Scope ............................................................7
2.2 Industry Architectures ...............................................................................9
2.3 Transactions ..............................................................................................13
2.4 Global Value Chains ................................................................................14
2.5 Industry Classification ............................................................................16
3. Theoretical Development ....................................................................................21
3.1 Theory .......................................................................................................21
3.1.1 Similarity ...........................................................................................21
3.1.2 Complementarity ..............................................................................24
3.2 Hypothesis Development .........................................................................26
3.2.1 The Determinants of Major Inter-firm Transactions ....................27
3.2.2 The Evolution of the Apparel Sector ...............................................31
4. Methodology ........................................................................................................35
4.1 Data ...........................................................................................................35
4.1.1 Data Preparation ..............................................................................35
4.1.2 Data of Apparel Sector .....................................................................36
4.1.3 Apparel Sector Roles ........................................................................37
4.2 Measures ...................................................................................................46
4.2.1 Dependent Variables .........................................................................46
4.2.2 Independent variables ......................................................................46
4.2.3 Control Variables ..............................................................................49
4.2.4 Descriptive Statistics and Correlation Matrix ...............................52
VI
4.3 Analytical Approach ................................................................................55
4.3.1 Fixed Effect Models ..........................................................................55
4.3.2 Random Effect Models .....................................................................55
4.4 Descriptive Statistics ................................................................................57
4.4.1 Data Records .....................................................................................57
4.4.2 Number of Unique NAICS Codes (4-digit level) per Year ............59
4.4.3 Number of Transactions and Value of Transactions ......................59
5. Results ..................................................................................................................61
5.1 Determinants of the Presence of Inter-firm Transactions ....................61
5.2 Determinants of the Magnitude of Inter-firm Transactions ................64
5.3 Comparison of Results .............................................................................66
5.4 Evolution of Apparel Sector ....................................................................71
6. Conclusion ............................................................................................................75
6.1 Contributions ...........................................................................................75
6.2 Implications ..............................................................................................77
6.3 Limitations and Future Research Directions ........................................80
References………………………………………………………………………….. 81
VII
List of Tables
Table 4.1 Central Roles in Apparel Sector………………………………………….….. 37
Table 4.2 External Roles with Frequencies of Appearance over 20………………….… 38
Table 4.3 Degree of Dependency of Sector Roles of Firms………………………....….. 46
Table 4.4 Summary of Variables and Measures……………………………………..….. 51
Table 4.5 Descriptive Statistics and Correlation Matrix…………………………..……. 53
Table 5.1 Logistic Regression of the Existence of Transaction on NAICS Proximity and
Dependency…………………………………………………………………..…………. 68
Table 5.2 GLS Regression of the Magnitude of Transaction on NAICS Proximity and
Dependency………………………………………………………………………...…… 69
VIII
List of Figures
Figure 3.1 Quadrant for Similarity, Complementarity, Intra-firm and Inter-firm
Relations…………………… …………………………………………………….……..18
Figure 4.1 Number of Unique Firms with Revenues…………………………….……....36
Figure 4.2 Transactional Network by NAICS Code (3-digit)…………………………....40
Figure 4.3 Transactional Network by NAICS Code (2-digit)…………………………....41
Figure 4.4 Transactional Network by Sector Role…………………………………….....42
Figure 4.5 Seller-Buyer Design Matrix Structure……………………………………..…43
Figure 4.6 Number of All Firms and Firms from Manufacturing, Service, and External
Sectors Every Five Years………………………………………………………….……..57
Figure 4.7 Trend per Five Years for Number of Unique NAICS Codes (4-digit level) of
Sellers, Buyers and Total Firms…………………………………………………..……...58
Figure 4.8 Trends for Number of Transactions and Value of Transactions…………..…..59
Figure 5.1 Proportion of Service Firms versus Manufacturing Firms………………..….71
Figure 5.2 Proportion of External Firms versus All Firms………………………….…...72
Figure 5.3 Proportion of Non-regular Transactions versus Regular Transactions…….....73
1
1. Introduction
Transactions are the most basic form of inter-firm relationship and are a fundamental
unit of economic analysis (Williamson, 1975). Within a sector, transactions in
intermediate markets serve to link and coordinate complementary activities across
firms (Jacobides & Winter, 2005; Baldwin, 2008). Research on transactions has
focused on the question of how a transaction will be governed. Transaction-cost
economics considers the firm’s decision as to whether to make or buy a particular
input (Williamson, 1975). Transaction costs are the costs associated with using
suppliers in the outside market rather than producing in-house. Opportunistic
transaction costs have been the focus of Williamson’s transaction costs economics,
while Baldwin (2008) places the emphasis on mundane transaction costs. Specifically,
opportunistic transaction costs plus mundane transaction costs are together equal to
the total transaction costs. Baldwin (2008) illuminates the importance of mundane
transaction costs, which are defined as “the costs of defining, counting and
compensating for things transferred”. The theory of mundane transaction costs
determines whether a transaction is possible, and argues that transactions are only
possible at thin crossing points between modules.
An inter-firm transaction is a basic business event and transactional relationships
impact firm prospects. Cohen and Frazzini (2008) show that customer-supplier
relationships generate significant co-movements in the underlying cash flows of
linked firms. A growing body of studies has investigated the impact of inter-firm
network structures on firm performance. Bonaccorsi and Giuri (2001) suggest that the
structural dynamics of the supplier industry depend on the ways in which inter-firm
transaction networks are created and distributed among vertically related supplier and
buyer industries.
A related and important question is where transactions are likely to take place. Given
2
a pair of firms, how likely is it that they will be engaged in a buy-sell relationship and
in which direction will the transaction be oriented? At an inter-industry level,
economists have reversed the question such that inter-industry transactions are used to
generate information on the structure of the economy. Input-Output (I-O) models
produce fundamental knowledge of the structure of the economy and are used to
reconcile government accounting systems. In some cases, global value chains (GVCs)
indicate the likely transactional relations between firms. For example, firms are
associated with relatively clear global value chain roles in sectors (e.g. automotives,
apparel) that produce physical products.
The Chief Economist of Foreign Affairs and International Trade Canada considers
global value chains (GVCs) to be of high importance and observes that “measurement
has probably been the most significant obstacle to developing a better understanding
of GVCs” (DFAIT, 2011). Similarly, in a recent report published by the Institute for
Research on Public Policy (IRPP, 2012), new data are identified as the primary
requirement for acquiring a better understanding of Canada’s role in GVCs.
The requirement for additional data arises as a consequence of the limitations of the
standard industry classification systems used by governments. These systems group
industries together into subsectors and sectors on the basis of similarities in the nature
of activities performed by firms in neighbouring industries. But value chains and other
self-organized networks of firms such as sectors and ecosystems are composed of
firms that perform complementary activities to collectively address the demands of
final user-customers. So while we know a great deal about relations between
competing firms that perform similar activities in the same industry, it is difficult to
obtain information about relations between firms in different industries that are
engaged in complementary supplier-customer relations. Therefore, the questions
remain. These include the following:
- What role do service providers play in global value chains?
3
- To what degree are firms outside the primary sector involved in the primary-sector
value chain?
- To what degree do transactions involve pairs of firms that have an
upstream-downstream relationship versus backward and cycled transactions?
- How stable are GVC relations over time?
It is of scientific importance to investigate why and where inter-firm transactions exist
because they help us to reveal the persistent pattern of inter-firm relations, and to
understand the structure of industry. I believe that large and persistent transactions
constitute the structure of the economy in the same manner as branches, not leaves,
capture the structure of trees. In the interest of developing a better understanding of
industry structure, I am conducting a study on the determinants of major inter-firm
transactions. In addition, it is of practical importance to know where inter-firm
transactions are likely to take place since transactions are firms’ fundamental
economic activities and are closely related to firms’ behaviours and level of
commitment to corporate social responsibility. Policymakers, firms and organizations,
and citizens should be aware of the impact of their economic activities on certain
critical issues such as human welfare and environmental welfare. For instance, the
disasters that took place in Bangladesh’s manufacturing industry have implications for
Western retailers and consumers in developed countries because their buying
behaviours have implications for labour and safety standards, including the
enforcement of building codes in Bangladesh (Strauss, 2013). Retailers and buyers,
though without official data, understand that their purchasing activities have
implications for other industries on which the retail industry depends; examples
include human-welfare issues in manufacturing industries or environmental issues in
raw materials and resource industries.
Governments have access to large quantities of data. But, because the data are
organized in an unhelpful way, governments cannot easily draw conclusions about
inter-industry or inter-firm relations from the data. They can only operate on the basis
4
of the common knowledge that consumers have, or on the basis of sensational news
reported by the public media. If the data that governments have were organized in
such a way as to make evident the relationships between upstream and downstream
economic activities, then we would have scientific knowledge regarding the impacts
of consumers’ purchases on human and environmental welfare. A better understanding
of inter-firm transactions would be of benefit to international trade policy (e.g., tariffs
and quotas), the improvement of labour and safety standards, and the development of
industrial policy.
My empirical focus will be on the apparel sector and this selection is justified for
several reasons. First, the apparel sector has experienced transformation in its supply
chain due to the introduction of the “fast fashion” concept and of “lean retailing”
strategies, along with the adoption of new information technologies (Abernathy et al.,
2000). Gereffi, Humphrey, and Sturgeon (2005) indicate that the apparel industry has
shifted from a captive value chain to a more complex relational value chain. Second,
the apparel sector can be considered to be one of the most global of all industries
(Dickerson, 1995; Gereffi, 1999). Globalization has facilitated industrial upgrading
that involves organizational learning to improve the position of firms or nations in
global trade networks (Gereffi & Tam, 1998). Thirdly, many U.S. economic analysts
hold the popular prognosis that U.S. apparel manufacturing is dying (Abernathy et al.,
1995). Tassey (2010) points out that the United States has allowed the manufacturing
sector to languish. The plight of the manufacturing sector not only is an “industry”
problem but is also a “value chain” issue. Finally, since my study is at the starting
point to investigate sector architecture, the apparel sector is suitable for testing a
pioneering theory due to its relatively low complexity and long tradition.
This study describes the nature and evolution of major inter-firm transactions in the
U.S. apparel sector. I believe that the fundamental patterns in the nature of inter-firm
relations persist, notwithstanding the strategic behaviours of different firms, the
presence and absence of certain firms over time, and the diversity in business
5
environments. Specifically, I will investigate the existence and magnitude of major
inter-firm transactions and will examine the structure and evolution of the U.S.
apparel sector over time.
My study sample relies on a unique longitudinal dataset of over 2,000 of the largest
transactional (buy-sell) relations between publicly traded firms in the U.S. apparel
sector over the last 35 years. This dataset allows me to characterize the determinants
and evolution of transactional networks. The data are available because the U.S.
Securities and Exchange Commission requires firms (including foreign firms) that are
publicly traded in the U.S. to report the percentage of their revenues that is
attributable to sales to a specific customer, in cases where those sales exceed 10% of
total revenues. I did not consider the minor inter-firm transactions because in general
large firms, rather than small firms, are more likely to construct and change the
structure of transactional networks. Moreover, small firms tend to follow the general
patterns and lack the capability to change the structure of industries.
This study only focuses on the apparel and apparel-related industries. Based on the
North American Industry Classification System (NAICS), I identify the following
subsectors: 313 (Textile Mills), 315 (Apparel Manufacturing), 316 (Leather and Allied
Product Manufacturing), 4243 (Apparel, Piece Goods, and Notions Merchant
Wholesalers), and 448 (Clothing and Clothing Accessories Stores); I also identify
observations where either the seller or the buyer operates in one of the
above-mentioned subsectors. Additional data on firms’ revenues are extracted from
Research Insight. The second dataset contains firm-year observations from 1992 to
2010.
My research questions are as follows:
- What factors predict the existence and magnitude of major inter-firm transactional
relationships?
- How does the U.S. apparel sector evolve over time?
6
My results clearly show that inter-firm complementarity predicts both the existence
and magnitude of major inter-firm transactions better than inter-firm similarity. My
thesis is the first study that responds in a scientific manner to the need to examine
major transactions between firms and that employs a theory that is replicable across
different sectors. This study also contributes to a deeper understanding of the apparel
industry by focusing on changes in the relative importance of manufacturing activities
versus service activities and in the involvement of firms from outside the apparel
sector.
The thesis is organized as follows. Chapter 2 provides a brief review of five strands of
literature. Then I develop my theories and derive hypotheses in Chapter 3. In Chapter
4, a description of my data and methodology is presented. In Chapter 5, findings and
results are presented and discussed. The last chapter discusses the contributions,
limitations, and implications of my study.
7
2. Literature Review
In this section, I provide a brief review of five strands of literature: 1) theories that
explain firm scope, 2) industry architecture, 3) transaction, 4) global value chains
(GVCs), and 5) industry classification. Each strand provides explanations for a better
understanding of industry architecture and its evolutionary path.
2.1 Theories that Explain Firm Scope
Vertical integration is a form of organizing from a firm-level perspective. Researchers
study vertical integration in terms of its theoretical issues and operational approaches.
It is relevant to my research because it deals with firm’s integration versus
specialization in terms of firm boundaries and it brings back the question of the
division of labor. Adam Smith (1776) argued that the division of labor is limited by
the extent of the market. The emergence of vertically integrated firms will redefine the
boundaries and generate re-division of labor among firms, and therefore transform the
architecture of industry as well as the whole economy. The overall trend is towards
specialization, although re-integration does occur.
Transaction cost economics. Transaction cost economics (TCE) theory particularly
focuses on the concept of vertical integration. As noted by Williamson (1975), the
main case in TCE is the issue of vertical integration. The core thesis of TCE is that
firm boundaries and the resulting levels of vertical integration are explained by
governance cost. Firms can choose to make the products that they need in-house or
buy them from outside suppliers. Transaction costs are the costs of governing make or
buy relationships. If government costs are high, the firms choose to make what they
need in-house, thus becoming more integrated. If the government costs are low, then
firms will buy from the outside market.
8
Transaction costs largely depend on the risk of opportunistic renegotiation
(Williamson, 1975). The risk of opportunism occurs when a firm involved in a
transactional relationship has to make dedicated and long-term investments to ensure
the relationship works well but this investment will only be supportive to this
relationship but not in other environments. The firm is likely to be ‘held up’ by the
firm from the other party when it makes asset-specific investments. Knowing the
potential risk, it is difficult for any firm to make such a durable but asset-specific
investment and get the valuable return of such a risky investment unless the firms on
the other side guarantee no risk of opportunism. However, firms behave
opportunistically, and are unlikely to guarantee no opportunism. So, the main way to
achieve specialized investments is vertical integration (Williamson, 1985). According
to the TCE theory, when asset specificity is high, and hence transaction costs resulting
from risk of opportunism are high, firms will chose to become vertically integrated.
TCE scholars consider market failures owing to asset specificity as the main reason to
explain firm’s decision of vertical integration (Williamson, 1985). Numerous studies
have found positive influence of asset specificity on vertical integration and asset
specificity has shown to be the most significant TCE-based determinant of vertical
integration (Gulbrandsen et al, 2009).
Resource-based theory. Resource-based theory focuses on the unique internal
resources of a firm’s sustained competitive advantage. Firms will tend to specialize in
activities that their capabilities offer some competitive advantages (Richardson, 1972).
Barney (2000) created a resource-based model of sustained competitive advantage
and suggested that sources of sustained competitive advantage are firm’s resources
that are valuable, rare, imperfectly inimitable and nonsubstitutable. RBV assumes that
firms develop by integrating activities that are related to existing activities
(Richardson, 1972). In addition, Richardson (1972) is very specific about the
relationship amongst the types of activities firms will do internally and the types of
activities that will be done by the firms with whom they transact. Winter (1988) points
out the notion that firms grow by vertical integration through developing ‘something
9
closely related’ instead of ‘more of the same’. He argues that where integration takes
place relies on the ‘degree’ of relevant resources and knowledge that firms already
had. Firms over time add activities that relate to some aspect of existing activities
(Teece et al, 1994). Based on RBV, as firms expand by vertical integration they grow
in the direction of something closely related to their existing resources, activities,
knowledge, experience and competency.
Following Winter (1988), the firm’s decision of integration should be assessed in
terms of the extent to which current resources and existing knowledge can be applied
to integrate a new activity. Coherence is a measure of relatedness proposed by Teece
et al (1994) and is exhibited by multi-product firms developing related lines of
businesses. It has shown that firms grow more diverse but that the level of coherence
between neighboring activities is keeping constant (Teece et al, 1994). Even though a
firm becomes more diverse by performing more new activities, its level of coherence
is maintained.
To summarize, TCE and RBV jointly provide the theoretical framework for issues of
vertical integration. TCE focuses on the firms’ trade-off between using outside market
resources and producing in-house, while RBV emphasizes firms’ unique capabilities
to create competitive advantage. According to TCE, firms become vertically
integrated due to the high transaction costs owing to high asset specificity and risk of
opportunism; whereas RBV explains that a firm’s decision of integrating new
activities is largely dependent on the relatedness of existing capabilities and the
capabilities required by performing new activities.
2.2 Industry Architectures
The concept of industry architecture literature examines the relationship between
firms and the industrial environment with regard to managerial strategy and the
10
performance of firms. As firms grow and evolve under the economic environment,
they may also contribute to the shaping and changing of industry architecture.
In the traditional Bain/Mason industrial organization paradigm, industry structure was
defined as the relatively stable economic and technical dimensions of an industry that
provided the context in which competition occurred (Bain, 1972). It is suggested that
industry structure (the degree of seller and buyer concentration, product
differentiation, and barriers to entry) determines the behavior or conduct of firms,
whose joint conduct then determined the collective performance of the firms in the
marketplace (Bain, 1968; Porter, 1981). The underlying notion is that industry
structure determines the behavior of firms, and hence determines the firm
performance, which was known as the structure-conduct-performance view. Scholars
who are concerned with firm performance must be concerned with industry structure.
The limitation of traditional industry organization paradigm is that it gives full
priority to the importance of the outside economic environment, but neglecting the
firms’ capabilities to change the structure. Brusoni et al (2011) explain what drove
particular industry participants to change the vertical structure, showing how firms
transform their industry environment, as opposed to merely reinforcing it.
Traditionally, industry structure is identified as number and size distribution of firms
with regards to concentration and barriers to entry. Industries were understood in an
atomistic sense as “a group of sellers or of close-substitute outputs who supply a
common group of buyers.” In that sense, industry structure refers to the nature of
product markets in terms of the concentration of seller and buyer, product
differentiation and barriers to entry (Bain, 1968; Porter, 1981). Homogeneous sector,
consisting of firms undertake similar activities, is the primary unit of analysis for
industry structure in an atomistic sense. Since inter-firm relations are widely observed
among various industry groups and products are supplied by large networks of
manufacturers and service providers, attention should be paid to the heterogeneous
sector that comprises firms engaged in complementary activities. From this relational
11
sense, industry structure is defined as the architecture of these relationships (Jacobides
et al., 2006).
Researchers who study industry architecture are interested in vertical integration and
vertical specialization. Jacobides (2005) provides an inductive theoretical framework
to explain how and why vertical disintegration happens, showing that transaction
costs are an incidental feature of industry evolution. He suggests that the real drivers
for vertical disintegration are benefits from specialization and potential gains from
trade. Jacobides and Winter (2005) argue that firms do not take their transactional
environment as given; rather they can operate on and change it. Transaction costs
theory does not address how industries or value chain structures evolve and does not
ask the question of whether firms can choose to make or buy (Jacobides, 2005).
Another work by Jacobides, Knudsen, and Augier (2006) qualifies Teece’s predication,
by positing that firm can create an “architectural advantage” in terms of high levels of
value appropriation without the need to engage in vertical integration and concludes
that firms can benefit from innovation by managing the industry’s architecture
carefully so they become the “bottlenecks” of their industry. Cacciatori and Jacobides
(2005) develop a framework to provide explanations that industries may shift to
vertical reintegration after long periods of vertical specialization. They observe that as
various groups shape and change the nature and boundaries of industries towards
vertical specialization, they create distinct knowledge bases in the industry.
Specialization of firms coupled with increasing distinct knowledge bases, firms in
such environment will suffer difficulties in coordination management across
boundaries and establishment of inter-firm relations. Reintegration is advanced by
firms seeking to protect their capabilities, through strategically changing their
institutional environment to create new all-in-one, integrated markets (Cacciatori &
Jacobides, 2006).
Besides vertical integration and disintegration, Jacobides and Billinger (2006)
examine permeable vertical architectures that are partly integrated and partly open to
12
the markets along a firm’s value chain. They indicate that firms can make, buy, and
ally by enabling upstream units to sell to other manufacturers and downstream unites
to buy intermediate products from other firms in the same sector. The firm therefore
designs a “vertically permeable boundaries”.
Luo and Magee (2010) develop metrics and methods to measure the degree of
hierarchy in transactional relationships among firms in two large industrial sectors in
Japan: automotive and electronics. The findings indicate that there is an absence of
hierarchy in transaction networks within electronic sector, because firms in electronic
sector exhibit two characteristics: 1) they participate in multiple stages of industry
value chains and hence are vertically integrated, but also 2) they purchase inputs from
and sell products to other firms in the sector (Luo et al., 2012).
While firms often trade off economizing on transaction costs versus access to
dispersed knowledge stocks and enhanced flexibility in marketing these important
governance decisions, many firms are partially integrated and simultaneously
outsource some activities (Harrigan, 1984). These firms seek to identify the most
effective balance in bother organizing alternatives to leverage their benefits and
mitigate their costs (Rothaermel et al., 2006). This organizing approach is labeled as
taper integration, following Harrigan’s theoretical contribution that firms taper
integration “when firms are backward or forward integrated but rely on outsiders for a
portion of their suppliers or distribution” (Harrigan, 1984). Bonaccorsi and Giuri
(2001) develop the argument that the long term structural evolution of an industry
depends on the evolution of a vertically-related downstream industry. They claim that
the evolutionary dynamics of the downstream industry, in terms of number of firms
and products, entry, exit and concentration, is transmitted to the upstream via the
structure of the network of vertical exchange relations (Bonaccorsi & Giuri, 2001).
13
2.3 Transactions
Transactions are the most basic form of inter-firm relationship and are a fundamental
unit of economic analysis (Williamson, 1975). Within a sector, transactions in
intermediate markets serve to link and coordinate complementary activities across
firms (Jacobides & Winter, 2005; Baldwin, 2008). Williamson (1985) defined a
transaction as “a transfer across a technologically separable interface”, though he did
not specifically explain what “technologically separable interface” is. To address the
question of why transactions occur in a particular place, Baldwin (2008) develops a
theory of location of transactions and the boundaries of firms in systems of production
(viewed as networks of tasks), drawing on modularity theory. Transactions are defined
as “mutually agreed-upon transfers with compensation, are located within the task
network and serve to divide one set of tasks from another” and are “designed to match
their locations” (Baldwin, 2008). Here, transaction is no longer the primary unit of
analysis and is more than a simple “transfer” indicated by Coase (1937) and
Williamson (1985), instead it is a “reciprocal exchange” (Baldwin, 2008) rooted in a
complex network structure.
Opportunistic transaction costs have been the focus of Williamson’s transaction costs
economics while Baldwin (2008) lays the emphasis on mundane transaction costs
which are defined as the “costs of defining, counting and compensating for things
transferred.” Specifically, opportunistic transaction costs plus mundane transaction
costs together equals to the total transaction costs. Drawing on modularity theory,
modules are loosely coupled at thin crossing points where transfers are fewer and less
complexity in comparison to transfers at thick crossing points within modules.
Therefore, placing transactions at module boundaries – thin crossing points –
generally achieve the lowest mundane transaction costs. When the network structure
is fixed and transactions are placed at thick crossing points, transaction designers
should make a tradeoff between mundane and opportunistic transaction costs to
14
achieve an optimal transaction. Also, the task network structure can be changed
through the process of modularization to make a thick crossing point become thinner.
Modularizations create new module boundaries with relatively low transaction costs
and thus make transactions feasible where they were previously impossible or very
costly (Baldwin, 2008). And the new module boundaries (new locations of
transactions) enable new entry of firms and upgraded competition, contributing to
changes of industry structure. There exist transaction-free zones where transfers occur
without the burden of costs. Baldwin (2008) indicates that the thin crossing points
between transaction-free zones constitute “breakpoints” that firms and industries
become partitioned.
2.4 Global Value Chains
In business management, the concept of the value chain was first introduced by
Michael Porter (1985). Value chain explains how value flows and consists of activities
that create value. According to Porter’s theory, firms should create sustainable
competitive advantage in order to survive and thrive. To achieve competitive
advantage, firms should develop a superior value chain or at least a more advanced
one than their competitors. At the firm level, the value chain represents a set of
activities and processes that firms are involved in. Those activities and processes
provide values to final customers in the form of product or service. But Porter’s value
chain was a series of intra-firm activities and concerned with the value flowing within
a single firm.
The current interpretation of a value chain refers to a series of sequential and
sometimes parallel activities conducted by multiple firms with complementary
capabilities. At an industry level, value chain is a physical representation of sequential
processes from design to distribution - starting with raw materials to manufacturing
and ending with final products and services.
15
The emergence of global value chains (GVCs) is a consequence of the global
integration of industries. A growing body of literature concerns the governance and
evolution of GVCs. GVCs can exhibit buyer-driven and producer-driven governance
patterns. Gereffi (2002) indicates that in the producer-driven GVCs, large
transnational manufacturers exert control over backward linkages with raw material
and subcomponent providers, and forward linkages into distribution and retailing. In
buyer-driven GVCs, large retailers, marketers and branded manufacturers play the
vital roles in setting up decentralized production networks (Gereffi, 2002). The
apparel industry is an example of buyer-driven global value chain. Three factors
determine the governance and transformation of global value chains: 1) the
complexity of transactions; 2) the ability to codify transactions; and 3) the capabilities
in the supply-base (Gereffi et al., 2005).The concept of GVCs contributes to
understanding three features of a particular industry or group of industries: 1) the
geography and character of linkages between tasks in the chain of value added
activities; 2) power distribution among firms in the chain; and 3) the role that
institutions play in structuring business relationships and industrial location (Sturgeon
et al., 2008).
Gereffi, Humphrey, and Sturgeon (2005) identify five basic types of value chain
governance as follows: market, modular, relational, captive and hierarchical
characterized by vertical integration. And their work indicate that apparel industry has
shifted from a captive value chain to a more complex relational value chain, and such
migration is mainly driven by increasing supplier competence. This notion is
consistent with findings from other studies that there has been a shift among apparel
suppliers in developing countries from providing basic assembly to offering more
flexible and efficient services, such as full-package production and lean or agile
manufacturing (Abernathy et al.,2000, 2006; Gereffi, 2001).
Imagine a simplest value chain raging from raw material suppliers, to manufacturers,
to product distributors and service providers, where each stage is associated with
16
higher value-added than its former stage. Conventional wisdom suggests that firms
who want to capture the most value, the best strategy is to position downstream in the
value chain and by the same token, countries who want to achieve economic
prosperity should encourage national firms to occupy the downstream positions and
leave the less value-added upstream positions to less developed countries. In the
United States, manufacturing sector has been offshore to less developed countries in
response to trends in creating and sustaining comparative advantage and efforts to
gain economic superiority on a global scale by capturing the downstream positions in
global value chains. However, there is skeptical view on the dominant growth strategy,
for instance, Tassey (2010) points out that United States has let the manufacturing
sector languish. The plight of the manufacturing sector is not only an “industry”
problem but also a “value chain” issue. He argues that the policy of prospering by
specializing in services ignores both the co-location synergies with manufacturing and
the fact that other economies are aggressively integrating forward into high-tech
services (Tassey, 2010).
In conclusion, global value chains capture value-flowing in the global economy. With
a thorough understanding of the direction of value flows, firms, organizations and
countries would tend to develop strategic polices to govern and adapt the value chains
for their advantage.
2.5 Industry Classification
Industry classification is an important criterion for researchers who aim to analyze
industry architecture. The works on identifying and investigating industry
architectures will be much easier if a standard industry classification is shared. Such
classification schemes as the US-based Standard Industrial Classification (SIC) and
North American Industry Classification systems have been introduced for a long time
and updated periodically. NAICS was developed under the auspices of the Office of
17
Management and Budget (OMB), and adopted in 1997 to replace the Standard
Industrial Classification (SIC) system (U.S. Census Bureau). However, studies on
industry classification have questioned the reliability and capability of those schemes
to correctly reflect industry architectures since those industry classification systems
have not changed correspondingly to the increasing complexity of industry activities.
Hicks (2011) examines the relationship of structural change in the economy and
industrial classification and indicates that the understanding of structural change is
compromised because scholars do not clearly articulate the limit of the classification
infrastructure that shapes empirical analysis. Dalziel (2007) proposes an alternative
systems-based approach to classification, taking the view that understanding industry
structure is not understanding competition among firms that conduct similar activities,
but understanding cooperation among firms that undertake complementary activities.
Dalziel (2007) distinguishes the traditional and current views of industry architecture
as “atomistic” and “relational”.
The conceptual foundations of firm classification schemes have two dimensions:
supply/demand orientation and structural assumption (Hicks, 2011). The supply
orientation shows a production basis while the demand orientation indicates a market
basis. From supply/production orientation, firms are grouped into a particular industry
based on the similarity of their production activities and processes. Meaures of
inter-industry relatedness capture the similarity of firm activities and resources. Bryce
and Winter (2009) developed a general inter-industry relatedness index that
contributes to empirical assessments that reply upon the concept of relatedness
between resources to characterize the direction of growth of the firms. Such a
relatedness index provides economic rationale for a firm participating in industry A to
perform and integrate activities of industry B. Whereas from demand/market
orientation, firms are clustered into an industry based on the complementarities of
their production processes and activities engaged in. In a production orientation, an
industry is a group of business establishments supplying similar outputs in the form of
product or service, and performing similar production activities; while in a market
18
orientation, an industry is a group of business establishments producing close
substitutes and involved in complementary activities.
The current dominant industry classification systems (e.g. NAICS) are from a supply
orientation and the underlying criterion for grouping is “similarity”. Building on the
“similarity” criterion, for example, apparel manufacturers are grouped in one industry
in that they are engaged in apparel-making activities, and the complementary firms
who supply textiles, sewing machines, information systems, transportation and
logistics services, and marketing and branding are grouped into other industries
because those firms are involved in different sets of activities. However, from a macro
level, all those firms collectively address consumer’s demand for clothing even
though they undertake different sets of activities.
There are three limitations of the traditional activities-based approach. First, they fail
to reflect industry structure. Specifically, they fail to provide indication of
relationships between firms since “the firms classified in a given category are no more
likely to have relationships with the firms in the neighboring categories, than they are
to have relationships with firms in other sectors” (Dalziel, 2007). Second, they group
firms that serve similar purposes/demands into different industries simply by
identifying their production activities. Dalziel (2007) suggests that competitors may
employ different technologies and conduct different activities to meet similar
demands, but they do not serve for fundamentally different demands. It should be
noted that supply-based systems may group firms that supply very different demands
together, but a demand-based systems may group firms that perform very different
activities together. Finally, the boundaries between manufacturers and service
providers have become blurred as a result of technological innovation. Since most
industries are doing business with the support of computer and information
technology, it is hard to distinguish the role of an IT firm between manufacturer and
service provider by just considering their production activities. The question of
industry classification is not so much how to group similar firms into industries but
19
how to group industries into sectors in a meaningful way.
Dalziel (2007) proposed a system-based approach to industry classification that takes
demand as the primary organizing principle, and the sector as the primary unit of
analysis, as opposed to the activities-based approach. She shifted the traditional focus
on ‘activities’ conducted by firms to ‘demand’ that multiple firms collectively respond
to and provided an application in communication equipment subsector to show that
the systems-based approach perform better reflects the industry structure and
inter-firm relations than the traditional North American Industry Classification system.
Dalziel (2007) points out that as a variety of complementary firms work together to
collectively address similar fundamental demands, the classification system’s unit of
analysis must accommodate firms that are different from one another, both in terms of
the nature of the specific demand to which they individually respond, and in terms of
the activities they undertake. This notion calls attention to the vertical sector that start
with fundamental demands, sequenced by interdependent layers of industries
collectively addressing the demands. The principle of the systems-based approach lies
in the dependency relationships. This means that final customers depend on service
providing firms to fulfill their needs and by the same token, service providers depend
on other complementary firms to supply input products/services. This hierarchical
dependency structure is consistent with the architecture of global value chain.
To summarize, the supply-oriented classification systems fail to consider the purpose
for which firms conduct activities and to whom and from whom they sell and buy
products/services. In addition, activity-based classification may blur the inter-firm
relations and isolate individual firms from the “network” by simply focusing on their
production activities. The conventional industry classification systems fail to capture
the division of labor and the network of firms. Industry structure has changed
correspondingly with the dense network of inter-firm relations, hence, industry
classification systems should change accordingly for the purpose to provide a holistic
view of economy for researchers and policy makers. Suggestions are given as follows:
20
1) considering heterogeneous sector rather than homogeneous industry as the primary
unit of analysis; 2) focusing on cooperation and complementarity rather than
competition and similarity among multiple firms; 3) taking a demand view to
understand the hierarchical dependency relationship of firms.
21
3. Theoretical Development
My theory is based on two concepts: inter-firm similarity and inter-firm
complementarity. In Section 3.1, I introduce these two concepts, and in Section 3.2, I
develop my hypotheses regarding the determinants of major inter-firm relations, and
evolution of the apparel sector over time.
3.1 Theory
Richardson (1972) considers industry as carrying out a large number of activities that
need to be carried out by firms with appropriate capabilities. He distinguished
activities as similar versus complementary. Activities that require the same capability
for undertaking are called similar activities while activities are complementary when
they represent different phases of a process of production and require coordination
(Richardson, 1972). The concept of inter-industry similarity mainly addresses the
“distance” between two industries (that the firms belong to) by calculating their
similarities (Teece et al., 1994; Bryce & Winter, 2009). The notion of inter-firm
complementarity refers to the inter-industry dependencies as indicated by input-output
tables, or as indicated by firms’ sector roles. In this study, inter-firm relatedness is
operationalized as the NAICS proximity, while inter-firm complementarity is
operationalized as the firms’ sector roles dependency (Dalziel, 2007).
3.1.1 Similarity
Several researchers have studied the degree to which firms’ engage in similar or
related activities. Especially for multi-product firms, they are active in many
industries that share common characteristics regarding production technologies
(Neffke & Henning, 2008). According to the prevalent industry classification systems
(e.g.NAICS), firms that perform similar activities are classified in the same industry.
22
The knowledge about similarity between industries is important to understand
economic transformation, firm growth and diversification, and firm performance
(Penrose, 1959; Rumelt, 1974). Given the importance of similarity or relatedness
between industries, research from strategic management area has developed different
measures to capture the degree of inter-industry relatedness between firms in different
industries.
Rumelt (1974) captures the concept of relatedness in his study on firm diversification.
Four broad categories of firm diversification are identified as single product firm,
dominant product firm, related product firm, and unrelated product firm. Three ratios
are used for classification as follows: 1) the specialization ratio which is defined as
the proportion of a firm’s revenues that is attributable to its largest discrete
product-market activity; 2) the related ration which is defined as the proportion of its
revenues that are attributable to the largest group of businesses that are related in
some way to one another; 3) the vertical ratio which is defined as the proportion of a
firm’s revenues attributable to all of the by-products, intermediate products, and final
products of a vertically integrated sequence of manufacturing operations (Rumelt,
1974). However, this procedure measures relatedness on a nominal level, only
allowing comparisons within group averages (Lien & Klein, 2009).
The majority of studies has been focused on firm diversification, but much less has
been directly emphasized on inter-industry relatedness. Recently, researchers have
constructed inter-industry relatedness indices based on the frequency with which pairs
of industries appear jointly in firm portfolios (Teece et al., 1994), technological
relatedness as measured by the technologically related of products produced by
different plants in Sweden (Neffke & Henning, 2008), or the joint industry
participation decisions of U.S. manufacturing firms (Bryce & Winter, 2009). Teece et
al (1994) develop a corporate coherence measurement to explain the similarities and
differences in the nature of coherence across firms and industries, drawing on the
survivor principle. They assume that “activities which are more related will be more
23
frequently combined within the same corporation”. Teece et al. (1994) count the
frequency of joint occurrences of pairs of industries in firm portfolio and standardize
this frequency to identify linkages between pairs of industries that exceed the
expected random assignments of industries. Following Teece et al.(1994), Bryce and
Winter (2009) upgrade the measurement and develop a general inter-industry
relatedness index which is a useful tool for assessing inter-industry relatedness in any
context requiring such measure (Bryce & Winter, 2009).They infer inter-industry
relatedness from the aggregate activity combinations of firms from both supply side
and demand side and such approach provides a strong resource-based measure of
relatedness because it reflects the unobservable ways that firms share resource among
industry activities (Bryce & Winter, 2009). However, the major limitation is that the
general relatedness index is currently computed only for manufacturing industries. It
is not applicable for studies that investigate industries from service provider sectors.
Neffke and Henning (2008) measure technological relatedness from a production
perspective between industries based on the dataset of products produced in plants in
all manufacturing industries in Sweden. However, they did not provide any real
quality assessment of their relatedness measure and their measure is also only for
manufacturing industries.
Relatedness based on SIC system is easily available. The 2-, 3-, and 4-digit levels of
SIC system are used as points on an underlying scale of relatedness, and arithmetic
values are assigned to the distances (Lien & Klein, 2009). However, the SIC-based
measures rely on the assumption that industries are homogeneous within category
levels which can be problematic when the breadth of the industry classifications
varies (Robins & Wiersema, 2003). More importantly, the SIC-based measurement
does not consistently reflect relationships among valuable resources in the ways that
firms actually combine them to create value (Bryce & Winter, 2009).
In general, the different measures can be classified into three categories: categorical
measures, RBV-based measures with regard to survivor principle, and SIC-based
24
measures. All of the studies that consider inter-industry relatedness employ similarity
in firm activities as their relatedness criterion.
3.1.2 Complementarity
There is a growing body of research that considers the importance of complementarity
in the firms’ knowledge resources and production activities. The landmark article of
Teece (1986) introduces the concept of “complementary assets” to explain the failure
of innovating firms to take the lion’s share of profitability. To successfully
commercialize an innovation, besides core technical know-how, there is need for
complementary capabilities. Such complementary requisites include services such as
distribution, marketing and branding, competitive manufacturing and after-sales
customer services. It is demonstrated that when imitation is easy, the significant
economic returns may accrue to the owners of such complementary assets, instead of
the innovators (Teece, 1986). Aspiring to extend Teece’s framework, Jacobides,
Knudsen, and Augier (2006) investigate how innovating firms can benefit from value
creation and appropriation with a focus on industry architectures, by treating
complementarity and mobility as two distinct components of co-specialization. The
study suggests that firms can strategically reshape the architecture of their sector by
encouraging competitions in complementary activities and restricting mobility and
competitions in their own segment (Jacobides et al., 2006).
A wide range of studies examine the inter-firm alliance formations and firm
performance with regard to complementarity. Arora and Gambardella (1990)
investigate the complementarities among strategies of external linkage of the large
biotechnology firms with different parties (e.g. universities, small/medium sized
research-intensive firms). They further indicate that the locus of innovation should be
a “network” of inter-organizational relations. Rothaermel (2001) focuses on the
inter-firm cooperation between incumbents and new entrants when the incumbents
have complementary assets within their firm boundaries that are critical to
25
commercialization. His study shows that, at the industry-level, incumbents exhibit a
preference towards alliances that leverage complementary assets over alliances that
focus on building new technological competencies. A study by Lin, Yang, and Arya
(2009) investigates firm performance as a consequence of firms’ alliance partner
selections. Combining the RBV and institutional perspectives, this study explores the
interactions of resource complementarity and institutional associations between the
firm and its alliance partners and demonstrates that having alliance partners with high
resource complementarities will boost firm performance.
Complementarity of firms’ knowledge resources is also crucial to firm performance.
Studies have investigated the impact of cross-business knowledge synergies in terms
of relatedness and complementarity on performance of multi-business firms. The
results indicate that synergies from product, customer, or managerial knowledge
relatedness do not improve firm performance, while synergies arising from the
complementarity of the three aforementioned types of knowledge relatedness
significantly improve both marked-based and accounting-based performance of the
multi-business firms (Tanriverdi &Venkatraman, 2005).
The aforementioned studies have operationalized complementarity in terms of firm
resources (e.g. knowledge resources, tangible assets, and R&D capabilities) to predict
firm performance and profitability, inter-firm alliances, and firms’ decisions on
vertical integration. In my study, complementarity is operationalized as the degree of
dependency of firms’ different sector roles to predict the likelihood of inter-firm
transactions.
26
-Firm diversification in terms of
product relatedness to explain the
performance differences of firms
(Rumelt, 1974, 1982)
-Coherence to explain firms’ growth
(Teece et al., 1994)
-Technological relatedness to
explain the composition and change
of a regions portfolio of
manufacturing industries (Neffke &
Henning, 2008)
- Relatedness index to predict entry
mode (Bryce & Winter, 2009)
-Complementary assets to
successfully commercialize an
innovation (Teece, 1986)
-Complementarities among
strategies of external linkage of
firms (Arora & Gambardella, 1990)
-Complementarities in terms of
firms’ knowledge resources
(Tanriverdi & Venkatraman, 2005)
- Predicting inter-firm relations in
the same industry (Schilling &
Steensma, 2001)
-Survivor-based relatedness to
predict firms’ decisions to enter new
markets (Lien & Klein, 2009)
-Incumbents prefer alliances that
leverage complementary assets
(Rothaermel, 2001)
-Encouraging competitions in
complementary activities can help
firms reshape the sector architecture
(Jacobides et al., 2006)
-Having alliance partners with high
resource complementarities will
boost firm performance (Lin et al.,
2009)
Figure 3.1 Quadrant for Similarity, Complementarity, Intra-firm and Inter-firm
Relations
3.2 Hypothesis Development
In the following part, I derive hypotheses for the major inter-firm transactions. The
hypotheses focus on the significance of inter-firm similarity and inter-firm
complementarity in predicting the existence and magnitude of a major transaction
between firms. In addition, I provide evidence on the nature and evolution of the U.S.
apparel sector on a macro level.
Complementarity Similarity
Intra-firm
Inter-firm
27
3.2.1 The Determinants of Major Inter-firm Transactions
Inter-firm Relatedness
Industries are comprised of groups of similar firms; therefore, inter-firm relatedness is
the relatedness of pairs of industries that firms belong to. One of the core concepts of
corporate strategy is inter-industry relatedness, which is considered important to
explain the performance differences of firms in terms of profitability (Rumelt, 1982),
to determine the direction of firm growth and diversification (Rumelt, 1974; Teece et
al., 1994), and to predict firms’ decisions to enter new markets (Lien & Klein, 2009).
Inter-industry relatedness takes “similarity” as the criterion for identifying major
inter-firm relations.
A large volume of studies have investigated the effects of inter-industry relatedness on
firm diversification, growth, entry mode, organizational choice, and acquisition and
merger performance. To show the relationship between diversification strategy and
firm profitability, Rumelt (1982) introduced categorical degree of relatedness to
capture the product diversity. He finds that the highest levels of profitability were
exhibited by firms diversifying primarily into related areas that share common core
skill and resources with current production activities. Bryce and Winter (2009) predict
the mode of entry of an expanding firm and find that the relatedness index is
significant in predicting the choice between acquisition and organic expansion. Neffke
and Henning (2008) measure technological relatedness between manufacturing
industries in Sweden and indicate that relatedness has significant explanatory power
for the composition and change of a regions portfolio of manufacturing industries.
Lien and Klein (2009) construct a survivor-based measure of inter-industry
relatedness to predict firms’ decisions to enter new markets.
Transaction costs economics provides theories on firms’ make or buy decision and
views the governance decision as dichotomous (Williamson, 1985). Another strand of
28
literature indicates that the choice for firms to integrate or to transact is not binary.
Parmigiani (2007) examines firms’ decision on simultaneously make and buy the
same good, instead of solely making or solely buying. Firms will employ concurrent
sourcing strategy because they desire to simultaneously monitor suppliers, produce
efficiently, and improve processes (Parmigiani, 2007). Recent studies (Parmigiani,
2007; Parmigiani & Mitchell, 2009) examine the concurrent sourcing strategy
employed by firms and suggest that concurrent sourcing resolves the coordination
versus flexibility quandary. Jacobides and Billinger (2006) introduces the concept of
“permeable vertical architectures” that are partly integrated and partly open to the
markets along a firm’s value chain, and such architectures enable firms to
concurrently use both internal and external suppliers and customers. Permeable
boundaries allow firms to selectively participate at multiple points along the value
chain (Jacobides & Billinger, 2006), and therefore engaging in tapered integration
(Harrigan, 1985). The concurrent sourcing strategy and the design of vertically
permeable boundaries relieve the tension between the traditional make-and-buy view
and modularity theory. Modularity theory predicts that firms tend to outsource
complementary activities to achieve flexibility (Sanchez & Mahoney, 1996; Baldwin
& Clark, 2000; Brusoni et al., 2001).
Several studies expect inter-firm relations between firms in the same industry, that is
to say, in the firms that undertake same activities. Schilling and Steensma (2001)
investigated three primary ways for firms to substitute tightly integrated activities
with loosely coupling activities and they are contract manufacturing, alternative work
arrangements, and alliances, with the purpose to test the significance of several forces
that drive the use of modular organization forms. The main argument for their study is
that there is a positive relationship between heterogeneity in the production process
(inputs and demands) for an industry and the use of flexible modular forms. However,
all their hypotheses are only partially supported, indicating either theoretical or
empirical difficulties. I believe the difficulties can be empirical because firms may be
less likely to contract and ally with firms in the same industry than with firms in other
29
industries. Tomlinson (2010) investigates relationship between inter-firm cooperative
ties and innovation in five UK manufacturing industries. This paper finds out that
vertical cooperative ties are significant in explaining firms’ level innovation
performance while horizontal cooperation between firms (i.e. cooperative with
competitors) does not appear to be significant in explaining innovation. Another study
by Lamprinopoulon and Tregear (2010) focuses on the inter-firm relations in small
and medium sized firm (SMEs) and explores the links with marketing performance.
The findings of their analysis suggest that the configuration of horizontal relationships
between producer SMEs has little bearing on marketing performance, unless
accompanies by strong vertical connections between key members of the SME cluster
and other actors in the supply chain. Therefore, I expect that transactions take place
between firms in the related industries, not between firms within the same industry.
Accordingly, I propose that a major inter-firm transaction is associated with inter-firm
similarity. But I expect an inverted-U relationship between inter-firm similarity and
major inter-firm transaction. Firms are more likely to transact with firms from similar
or related industries, but will be less likely to transact with firms from unrelated
industries where there is no supply-demand relationship or with firms from the same
industry where there is more competition than cooperation.
Hypothesis 1:
There is an inverted-U relationship between inter-firm similarity and the existence and
magnitude of major inter-firm transaction.
Inter-firm Complementarity
Even though “complementarity” was widely examined by studies in the areas of firm
boundaries, profitability of innovators, industry architectures, strategic alliances, and
firm performance, there is no applicable measure of complementarity. For inter-firm
complementarity, it is difficult to systematically identify complementarity and even in
30
vertical value chain, there is absence of general means of identification.
One approach to identify the inter-firm complementarity is to use Input-Output (I-O)
tables. I-O analysis was first proposed by the Nobel Prize laureate Wassily Leontief
and describes inter-industry transactions as a proportion of total industry inputs and
outputs. I-O analysis tracks the interdependence among various producing and
consuming sectors of an economy (BEA, 2012). Input-Output tables indicate
inter-industry and therefore inter-firm complementarity. The bigger the proportion of
industry i’s input that is procured from industry j, the more likely that there will be a
buy-sell relations between firms in industry i and j. However, the I-O tables are not
sufficiently fine-grained to indicate the likelihood of inter-firm transactions across all
industries. In addition, the I-O tables are produced infrequently and usually with a lag
of thirty months for national tables, failing to keep up with the recent trends in several
industries and the changing economy.
Another approach is to use the concept of “sector role” in value chain and
operationalize complementarity as firms’ sector role dependencies (Dalziel, 2007).
From a demand perspective, there exist dependency relations between sector roles.
For instance, as service providers depend on manufacturers to provide inputs, by the
same token, manufacturers depend on material providers to supply materials. Sector
roles are defined on the basis of complementarity in activities. Within the vertical
structure of value chain, inter-firm relations will be hierarchically ordered from firms
providing services to final customers at the highest level, firms manufacturing
products at intermediate level, and firms supplying raw materials at the lowest level.
Consistent with the structure of global value chains, sector roles are ordered from
upstream sellers to downstream buyers. Dalziel (2007) proposed a theory-based
approach to identify the most important inter-industry complementarity. Her concept
of sector role is based on the modularity theory (Baldwin & Clark, 2000) that
identifies two sector roles – systems integrator and component supplier. She identifies
sector roles and distinguishes between service and manufacturing subsectors and
31
between central firms and complementary firms.
Previous studies have investigated the impact of a firm’s sector role on firm
performance and profitability. Sector role is important to explain why firms in the
same sector may experience the same environmental conditions differently (Dalziel &
Zhang, 2010). Accordingly, I propose that the existence and magnitude of a major
transaction between two firms is positively associated with their inter-firm
complementarity (dependency relation between sector roles).
Hypothesis 2:
There is a positive relationship between inter-firm complementarity and the existence
and magnitude of major inter-firm transaction.
3.2.2 The Evolution of the Apparel Sector
Recent studies have described inter-firm network structure and tried to explore
variation in network structure across industries. Rosenkopf and Schilling (2007)
compare alliance networks in 32 industries and demonstrate substantial differences in
network structures. They identify three types of networks: 1) disconnected with low
connections; 2) spider-webs with a high level of connections; and 3) hybrids with a
moderate level of connections. They further explain this variation by referring to
technological uncertainty and dynamism, product modularity and architectural control
that characterize the industries. Luo (2009) measures the degree of hierarchy of
transaction networks among firms in Japanese automotive and electronics sectors and
the empirical results show that the electronics sector exhibits a significantly lower
degree of hierarchy than the automotive sector, due to the presence of many
transaction cycles. It is worth reporting that the cycles in the electronics industry
disappear if the largest firms are taken out of the sample.
Since my data contains information at an industry level and the NAICS codes enable
32
me to identify the four apparel sector roles1, I expect to provide evidence on structure
of the U.S. apparel sector on a macro level. My first assumption is that there will be a
shrinking proportion of manufacturer firms involved in transactions in the U.S.
apparel sector. This expectation is consistent with Tassey’s (2010) view that the U.S.
outsources and offshores its manufacturing activities to less-developed countries
because U.S. firms are more willing to capture the downstream positions that take the
most value-added to gain economic superiority. The second expectation is that there
will be an increase in the diversification of multiple players that serve as external
firms from non-apparel industries and undertake complementary activities to apparel
firms, because nowadays products and services are provided by networks of firms.
Abernathy et al (1995) provide empirical results that apparel supplier have made
increasingly investment in information technologies, distribution systems, and other
associated services.
Hypothesis 3:
Over time, there will be a shrinking proportion of manufacturing firms relative to
service providers involved in major transactions in the U.S. apparel sector.
Hypothesis 4:
Over time, the proportion of major transactions that involve firms outside the apparel
sector will increase.
Luo and Magee (2011) present a metric and technique to quantitatively assess the
extent to which self-organizing directed networks exhibit a flow hierarchy. In the
appendix, they differentiate all the links of a network into four different types:
1) Regular: the link connects from anode on a pre-defined low level (i) to a node on
its adjacent higher level (i-1)
2) Level-Skipping: the link connects from a node on a pre-defined lower level (i) to a
node on a level (j) higher than its adjacent higher level (i-1), i.e. j < i-1; 1 Apparel sector roles comprise of Textile provider, Apparel manufacturer, Wholesaler, and Retailer.
33
3) In-layer: the link connects between nodes on the same level (i)
4) Backward: the link connects from a node on a predefined high level (i) to a node
on a lower level (j), i.e. i < j
Following Luo and Magee (2011), I classify the transactions in the apparel industry
into four types: 1) Regular (one-level up), 2) Level-skipping, 3) Same-level, and 4)
Backward. My expectation is that regular transactions will be more likely to occur
than the other types of transaction. For example, within the apparel sector, a
preponderance of links will be found between textile providers and apparel
manufacturers or between apparel manufacturers and retailers while relatively less
links will be found between textile providers and retailers. The directions of links are
expected to be consistent with the value flows in global value chains. The majority of
links will start from upstream sector roles and travel to downstream sector roles.
However, there can be backward links and cycles because of the existence of
vertically and tapered integrated firms and the fact that firms develop permeable
boundaries and employ concurrently sourcing strategy. While vertical disintegration
may be the long-term trend; over short time periods both the bicycle components
industry (Fixson & Park, 2008), and the British building industry (Cacciatori &
Jacobides, 2005) experienced vertical re-integration as a consequence of strategic
choices made by firms. Firms are partially integrated and concurrently sourcing in
order to achieve the most effective balance in both organizing alternatives to leverage
benefits and mitigate costs (Rothaermel et al., 2006). Harrigan (1984) defines this
organizing approach as taper integration that firms are backward or forward integrated
but also relying on outsiders for a portion of their supplies or distribution. Jacobides
and Billinger (2006) conducted a longitudinal study of a major European apparel
manufacturer and documented its redesign into permeable vertical architecture that
allows firm units to both make and buy, and transfer downstream or sell. It is
indicated that increased permeability enables more effective use of resources and
capacities, better matching of capabilities with market needs, and benchmarking to
34
improve efficiency (Jacobides & Billinger, 2006).
Hypothesis 5:
Over time, the proportion of the level-skipping, same-level and backward transactions
relative to transactions between vertically related pairs of industries will increase.
35
4. Methodology
In this section, I first provide information about my data and identify my study sample to
test hypotheses. In the following part, I introduce my analytic approaches and measures
of dependent variables, independent variables and control variables. The last part
presents the results of descriptive statistics.
4.1 Data
I am fortunate to have access to a dataset of over 60,000 observations of major
transactional relations in the U.S. from 1976 to 2010. Cohen and Frazzini (2008) have
used this dataset for their award-winning research in finance area to explore the
customers’ impact on suppliers. Hertzel et al (2008) also use this dataset to study
bankruptcy spillover effects on the stock prices of customers, suppliers, and strategic
alliance partners. Even though this dataset provides a holistic view of transaction
networks over time, it has not been used to investigate the transaction networks’ structure.
The data depends on the fact that the U.S. Securities and Exchange Commission requires
firms (including foreign firms) that are publicly traded in the U.S. to report the
percentage of their revenues that is attributable to sales to a specific customer, in case
where those sales exceed 10% of total revenues.
4.1.1 Data Preparation
The data preparation consists of two major tasks: 1) Standardizing company names, and
2) Identifying customer industry NACIS codes. The seller names are identified, so the
task of standardization is to standardize the buyer names. The first criterion is to remove
observations where the customer types are not “company” and where customer (buyer)
names are incomplete. Because firms report their buyer names by various formats such as
a full name, a full name plus Corp, or an abbreviation, a same company can be identified
36
by different company names. Therefore, with the help of a fuzzy string name-matching
algorithm to compare names, firm names with different reporting formats are identified
as the same firm in the whole dataset. The second task is to identify the missing NAICS
codes for buyers. The first step is to identify the unique buyer names and then to connect
those unique firm names with the dataset of Research Insight to find the corresponding
NAICS codes.
The result dataset includes 61,036 data records with gvkey codes for sellers, seller names,
NAICS codes for sellers, customer identifiers, unique identifier for buyers, buyer names,
NAICS codes for buyers, company type, geographic area code, geographic area type,
annual sales, segment identifiers, segment type, and reporting date of data records.
4.1.2 Data of Apparel Sector
Since my main interest lies in the apparel sector, I filter the data from the original dataset
that contains 61,036 data records. Each record is one row, presenting information (e.g.
name, NAICS Code) on a given seller, the corresponding buyer, and the transaction that
links the two. Of my research interest, data are extracted only from the apparel and
apparel-related industries. Therefore, based on North American Industry Classification
System (NAICS), I identify subsectors as follows: 313 (Textile Mills), 315 (Apparel
Manufacturing), 316 (Leather and Allied Product Manufacturing), 4243 (Apparel, Piece
Goods, and Notions Merchant Wholesalers) and 448 (Clothing and Clothing Accessories
Stores), and identify observations where either the seller or the buyer operates in areas of
the above selected subsectors. Same as my approach, Baldwin and Clark (2000)
identified new emerging sub-industries based on the SIC 4-digit codes to investigate the
changing structure of the computer sector.
As a result, the newly-extracted apparel industry dataset contains 2431 unique
observations in total from 1976 to 2010. Through filtering the “firm names” column, we
identify 277 different sellers and 135 different buyers. In case that there are overlapping
37
firms appearing in both sellers and buyers, we identify 381 unique firms while there are
31 firms serving as both a seller and a buyer.
As the above-mentioned dataset is only about the inter-firm transaction relationships,
further data is expected to indicate firm attributes. I extracted the data about firm
revenues from Research Insight using the unique firm names. The new dataset is about
3142 records of firm information from 1992 to 2010 and the earliest data records in
Research Insight started from 1992. I fail to find all the unique firms identified from the
original apparel dataset because some firms may be acquired or closed before the year
1992, resulting the missing and changing of firm names in the dataset of Research Insight
software. Totally, there are 289 unique firms in the dataset of firm information while the
total unique firms identified from my original apparel dataset are 381. Figure 4.1 shows
the number of unique firms having information about revenues from 1992 to 2010. It is
worth mentioning that there are different firms for each year and it is rarely possible for a
particular firm appearing for the whole period.
Figure 4.1 Number of Unique Firms with Revenues
4.1.3 Apparel Sector Roles
I identify four major sector roles within the apparel sector: 1) Textile provider, 2) Apparel
manufacturer, 3) Wholesaler, and 4) Retailer. For sector roles, there is a “dependency”
158
217 207 212 225 219
206 196 186 177 164 157
145 131 120 113 107 102 98
0
50
100
150
200
250
Number of firms (having revenues data)
38
(Dalziel, 2007) pattern that retailers depend on wholesaler to supply inputs, wholesalers
depend on apparel manufacturers to provide inputs, and apparel manufacturers rely on
the textile providers to offer inputs. From a demand perspective, firms are positioned by
the relation of dependency.
In an industry level and based on the NAICS codes (4-digit level), textile provider
includes business establishments from industry groups of 3131, 3132, 3133, and 3141;
apparel manufacturer contains business establishments from industry groups of 3151,
3152, 3162, and 3169; the wholesaler is the industry group of 4243; and the retailer
contains business establishments from industry groups of 4481, 4482, 4483, 4511, 4521,
4529, and 4541. Other left firms that undertake facilitating activities are categorized as
external roles and they are from non-apparel sectors. External firms from other sectors
(e.g. lessors of real estate, electronics and appliance stores) facilitate the operation of the
apparel sector, but they are not necessary for the apparel sector. In Table 4.1 below, I
present the classification of NAICS (4-digit level) industry groups in the apparel sector
according to their sector roles. Table 4.2 provides the information for non-apparel
industries that appear more than 20 times.
Table 4.1 Central Roles in Apparel Sector
Apparel sector roles (Associated with 4-digit NAICS codes)
Textile provider
3131:Fiber, Yarn, and Thread Mills
3132: Fabric Mills
3133: Textile and Fabric Finishing and Fabric Coating Mills
3141: Textile Furnishing Mills
Apparel manufacturer
3151:Apparel Knitting Mills
3152:Cut &Sew Apparel Manufacturing
3162:Footwear Manufacturing
3169:Other Leather and Allied Product Manufacturing
Wholesaler
4243: Apparel, Piece Goods, and Notions Merchant Wholesalers
Retailer
4481:Clothing Stores
39
4482:Shoe Stores
4483:Jewelry, Luggage, and Leather Goods Stores
4521: Department Stores
4529: Other General Merchandise Stores
4541: Electronic Shopping and Mail-Order Houses
Table 4.2 External Roles with Frequencies of Appearance over 20
External roles Frequency
3256: Soap, Cleaning Compound, and Toilet Preparation Manufacturing 25
3342: Communications Equipment Manufacturing 23
3361: Motor Vehicle Manufacturing 34
3399: Other Miscellaneous Manufacturing 26
4461: Health and Personal Care Stores 21
4511: Sporting Goods, Hobby, and Musical Instrument Stores 39
5311: Lessors of Real Estate 87
5416: Management, Scientific, and Technical Consulting Services 21
The external roles exist due to the fact that firms in the apparel sector have transactions
with firms from external sectors (e.g., Lessors of real estate, and Electronics and
appliance stores). The justification for identifying sector roles and external roles can be
explained that a complete structure of the apparel sector cannot be outlined by only
looking at groups of business establishments within the apparel sector especially when
the boundary between different industries is becoming blurry and the vertical cooperation
among complementary firms is overwhelmingly intense.
4.1.4 Transactional Networks in Apparel Sector
To have a better understanding of the transactional (buy-sell) relations in apparel sector, I
present graph visualizations created by Netdraw to show the transactional networks of
sellers and corresponding buyers. A graph presenting the information among relations
can efficiently provide visual evidence on the overall structure of networks. Netdraw is a
visualization tool from the social network analysis software package UCINET and it
allows graphic representation of networks including relations and attributes. Network
analysis uses graphic display that consists of nodes to represent actors and lines to
40
represent ties or relations (Borgatti, Everett, & Freeman, 2002). Each node has a set of
attributes and each tie can be directed by the arrow. Coloring the nodes by attributes is a
powerful way of exploring general patterns of network.
To construct my transactional network, I use nodes to represent unique firms involved in
major transactions in apparel sector and lines to represent the existence of a transactional
relation. The direction of a buy-sell relation is shown by the arrow. According to value
chains, firms are ordered from textile providers, manufacturers, wholesalers, to retailers
from upstream to downstream. I create a single node named “universal customer” and
create ties that link retailers to this “universal customer”. By doing so, all retailers will be
positioned near the center. I also include the transactional relations between external
firms2 (non-apparel firms) and apparel firms. In the following, I show the graph
visualizations of transactional network identified by node attributes, namely, NAICS
code (2-digit and 3-digit) and sector role. Different NACIS codes and different sector
roles are distinguished by the unique colors as presented in graphs.
Figure 4.2 shows the transactional network indentified by the node attribute: NAICS
code (3-digit). The dark pink in the center is the universal customer that connected with
all retailers. Overall, nodes by different colors are sparsely scattered. The color with the
most frequency of appearance is the sandwash pink that represents NAICS code 315,
reporting that the majority of apparel firms involved in major transactions are from
NAICS subsector 315 (apparel manufacturing). Besides, it is difficult to visually identify
any particular patterns of interest from the messy graph.
2 External firms are from non-apparel industries but are involved in transactions in apparel sector.
41
Figure 4.2 Transactional Network by NAICS Code (3-digit)
Figure 4.3 shows the transactional network identified by the node attribute: NAICS code
(2-digit). The dark blue is the universal customer created to link with all retailers. The
colors with the most frequencies of appearance are the orange that represents the NAICS
sector 31 (Manufacturing), the lemon and the red that represents the NAICS sector 44-45
(Retail Trade). Generally, there are certain clusters in the central part of the network in
terms of the relations between firms from manufacturing sector and firms from retail
trade sector. But it is hard to clearly indentify general patterns.
42
Figure 4.3 Transactional Network by NAICS Code (2-digit)
Figure 4.4 presents the transactional network indentified by firm attribute sector role.
Here, I classify four different sector roles, namely, textile provider, apparel manufacturer,
wholesaler, and retailer in apparel sector, and I also include external role to represent
firms from non-apparel industries. The dark green in the center is the universal customer
created. The red nodes are retailers that are closely circled around the universal customer.
The lemon nodes are apparel manufacturers and the pink nodes are textile providers. The
three dark blue nodes are wholesalers that appear rarest in the apparel transactional
network. The grey nodes represent external firms and they circle the periphery of the
transactional network. Overall, the different clusters by color can be identified obviously
from the graph. And there are tiers by color from the center to the edge of the
transactional network, which is consistent with the successive stages of value chains. The
graphs collectively indicate that sector role concept better captures the transactional
relations between firms than NAICS-based measure.
The network analysis yields a visualized presentation of the structure of apparel sector. It
provides visual support to the proposition that firms’ sector role dependency will predict
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29 30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 61
62
63
64
65
66
67
68
69
70
7172
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
43
major inter-firm transactions. In addition, my results further imply that the theoretical
sector role dependency concept will better predict the inter-firm relations than the
empirical relatedness measure based on NAICS.
Figure 4.4 Transactional Network by Sector Role
Figure 4.5 presents the seller-buyer design matrix structure (DMS). The X axle and Y
axle are all unique firms ranging from retailers, wholesalers, apparel manufactures, to
textile providers from upstream to downstream. Retailers are taking the numbers from1
to 33. The following numbers 34 and 35 are wholesalers. Apparel manufacturers are
taking the numbers from 36 to 190, and the rest numbers from 191 to 214 are textile
providers. Each pink dot represents the existence of a major inter-firm transaction. The
DMS graph reports the preponderance of the one-level skipping transactions from
apparel manufacturers to retailers. There are also several level-skipping transactions from
textile providers to retailers. Also, there is presence of same-level transactions between
apparel manufacturers themselves and between retailers themselves. The DMS graph
suggests the presence of vertical or tapered integration of firms.
44
Number 1-33 34-35 36-190 191-214
Sector role Retailer Wholesaler Manufacturer Textile provider
Figure 4.5 Seller-Buyer Design Matrix Structure
4.1.5 Study Sample
The study sample for testing the determinants of the magnitude of major transactions
includes 599 dyads of firms that have major transactions with each other in U.S. apparel
sector from the year 1992 to 2010. It is a panel data contains observations both on
firm-level attributes and dyad-level attributes.
To address the determinants of the presence of major transactions, I need to construct a
transactional network and compute all the possible pairs of firms that have transactions
with each other. In terms of the existence of major inter-firm transactions, in essence, the
empirical component is a transactional network of linkage establishments between firms.
Buyer
Seller
45
I model the possibility that a firm will have transactions with another firm and construct
adjacency matrixes to capture the relationships between firms in a transactional network.
I also include the reverse-directed dyads, that is, firm A sells to firm B is different from
firm B sells to firm A, because the transactions are directed in my study. Such a broad
definition of the dyads of firms at risk of having transactions is essential to producing
unbiased results, though; including a large number of dyads of firms that never have
transactional relations may create their own set of biases (Gulati, 1995).
Similar to my study on transactional networks, previous social network studies of tie
formation analyze every possible dyad (Podolny, 1994; Gulati, 1995; Stuart, 1998),
because they do not have any additional criteria to determine which dyads are likely to
exist and which dyads are not. Such strategy has been criticized for not accounting for
non-independence problem as each firm enters the analysis for numerous times
(Sorenson & Stuart, 2001). In addition, this strategy has another practical problem, that is,
to compute all the possible dyads is burdensome. For example, I have to create an
adjacency matrix with over one hundred thousand cells for consideration of all the
possible pairs. Sorenson and Stuart (2001) solved the problem by creating a matched
sample of potential dyad ties that did not exist while including all dyads of relations that
appear in the data. However, their results can be biased because the proportion of positive
outcomes in the sample does not match the positive outcomes in the population
(Sorenson & Stuart, 2001). In my panel data, most dyads of firms appear for certain
periods and then disappear in later years. There is no need to produce a redundant matrix
using all the unique firms appearing in the panel data over 18 years. Therefore, my
solution is to compute all the possible dyads of firms year by year, that is, to create
annual matrix (cell*cell calculation) only using unique firms for each year. My approach
substantially reduces the numerous entries of the same firms and the sample size is only
over 20,000 as opposed to 100,000. The only problem is that my sample is an unbalanced
panel data that may entail some computational and estimation issues, but most software
packages (e.g.STATA) are able to handle both balanced and unbalanced data.
46
4.2 Measures
In the following section, I introduce the dependent variables, independent variable and
control variables in this study. In Table 4.4, I summarize the variables and measures.
4.2.1 Dependent Variables
My dependent variable is a measure of inter-firm transaction; I envision two alternatives
to this measure: 1) Existence of a major inter-firm transaction; 2) Magnitude (value) of a
major inter-firm transaction. The first alternative is a binary variable while the second
one is a continuous variable.
Existence of a major transaction
The first dependent variable is the existence of the link (transaction relationship) between
two firms. It is measured by the existence of a link between a seller firm and a buyer firm.
It is reported as a binary variable that takes the value of 1 when there is existence of a
link between two firms and 0 otherwise.
Magnitude of a major transaction
The second dependent variable is the size of the link measured by the value of a major
transaction between firms. The second dependent variable is a continuous variable. It is
reported as the magnitude of transaction between firm i and firm j.
4.2.2 Independent variables
Inter-firm similarity
Inter-firm similarity is measured by the proximity of 4-digit NAICS codes of two firms
that have transactions with each other. This is a categorical variable taking the value of 3
when the 4-digit NAICS codes of seller firm and buyer firm are totally the same, or
47
taking the value of 2 when the 4-digit NAICS codes of seller and buyer are the same in
their first three digits, or taking the value of 1 when the 4-digit NAICS codes are the
same in their first two digits, and taking the value of 0 when the 4-digit NAICS codes of
seller and buyer are different. It is worth note that in NACIS classification, the 2-digit
codes 31-33 (31, 32, and 33) are all classified under manufacturing sector and the 2-digit
codes 44-45 are both classified under retail trade sector. In addition, the broadest valid
NAICS codes are at 2-digit level. Therefore, for instance, the proximity of NAICS codes
between a firm in industry 4481 and a firm in industry 4521 is taking the value of 1
instead of 0.
Inter-firm complementarity
Inter-firm complementarity is measured by the degree of dependency of firms’ sector
roles. This is a categorical variable taking the value of 1 when there is direct dependency
between sector roles, or taking the value of 0.5 when there is one-level indirect
dependency between sector roles, or taking the value of 0.25 when there is two-level
indirect dependency, taking the value of 0 when there is absence of dependency between
sector roles, taking the value of -1 when there is a direct backward link from downstream
to upstream, taking the value of -0.5 when there is one-level indirect backward link,
taking the value of -0.25 when there is two-level indirect backward link between two
firms. The direction of dependency relation between sector roles is supposed to from
upstream seller to downstream buyer, but there are also reversed-directional links
between firms partially due to the presence of firm integration. The following table 4.3
presents the classification of the seven different categories.
Table 4.3 Degree of Dependency of Sector Roles of Firms
Category Value Definition Example
48
Direct 1 The dependency relation is from
upstream to downstream.
There is only one level-up
between the two sector roles.
Textile
Provider
Apparel
Manufacturer
Indirect
(One-level
skipping)
0.5 The dependency relation is from
upstream to downstream.
There is one-level skipping
between the two sector roles.
Apparel
Manufacturer
Wholesaler
Retailer
Indirect
(Two-level
skipping)
0.25 The dependency relation is from
upstream to downstream.
There is two-level skipping
between the two sector roles.
Textile
Provider
Apparel
Manufacturer
Wholesaler
Retailer
Absence 0 There is no dependency relation
between the sector roles.
There is a link in the same layer
of sector role.
Apparel
Manufacturer
Apparel
Manufacturer
Backward
(Two-level
skipping)
-0.25 The direction of dependency
relation is reversed.
There is a backward link from
downstream to upstream.
There is two-level skipping
between the two sector roles.
Wholesaler
Apparel
manufacturer
Retailer
Textile
provider
49
Backward
(One-level
skipping)
-0.5 The direction of dependency
relation is reversed.
There is a backward link from
downstream to upstream.
There is one-level skipping
between the two sector roles. Wholesaler
Apparel
manufacturer
Retailer
Backward
(Direct)
-1 The direction of dependency
relation is reversed.
There is a backward link from
downstream to upstream.
There is only one-level up
between the two sector roles.
Textile
Provider
Apparel
Manufacturer
4.2.3 Control Variables
Besides independent variables, I control for a number of variables to ensure that my
findings are not caused by other factors that might affect the existence and the magnitude
of major transactions. The first concern is that the existence and magnitude of
transactions can be strengthened or weakened by the variations of firms’ attributes. Firm
size indicates its degree of success in the marketplace (Gulati, 1995). It is assumed that
firms with large sales and high profitability are more likely to have transactions with
other firms and much easier to attract potential transactional partners. Therefore, I control
for firm revenues of both sellers and buyers. Besides, I also control for the in-degree and
out-degree firm attributes in a major transaction.
Buyer firm revenues
Buyer firm revenue is measured by the revenue in year i for a particular buyer firm that
engage in a major transaction in year i. This explanatory variable is continuous.
Seller firm revenues
Seller firm revenue is measure by the revenue in year i for a particular seller firms that
50
engage in a major transaction in year j. It is a continuous control variable.
Number seller outgoing links
I measure the outgoing links by calculating the number of transactions that a sellers have
with all other buyers in a particular year. This firm-level variable gauges the degree of a
seller firm’s capability to have major customers in a given year. By the number of
outgoing links, I control the effect that firms with lots of customers will more likely to
have a link to a specific downstream industry.
Number buyer incoming links
I measure the incoming links by calculating the number of transactions that a buyer have
with all other sellers in a certain year. This firm-level variable captures the momentum of
a buyer firm’s capability to have major transactions with seller firms. By controlling the
number of incoming links, I control the effect that firms with lots of suppliers will more
likely to have a link from a specific downstream industry
Value seller outgoing links
I measure the outgoing value by summing the total value of all transactions that a seller
have with all other buyers in a certain year. This firm-level variable captures the effect
that a firm with a great number of outgoing values is more capable of having transactions
with other firms. Alternative interpretation is that this variable controls for issues relating
to the finite capability of a seller firm to have transactions with buyer firms.
Value buyer incoming links
I measure the incoming value by summing the total value of all transactions that a buyer
have with all other sellers in a certain year. This firm-level variable captures the effect
that a firm with a large amount of incoming values is more likely to have transactions
with other firms. Another interpretation is that this variable controls for a buyer firm’s
finite capability to have transactions with suppliers.
51
The second concern is that the prior transactions between two firms may lead to future
transactions. Firms tend to have transactions with their previous transactional partners
due to a variety of reasons such as established mutual trust and low communication cost.
To control for this factor, I compute two variables to capture the impact of prior links
between two firms on having subsequent transactions in later years. In addition,
including these two variables also controls for dyad-level heterogeneity.
Previous transaction
Previous transaction is a dummy variable that indicate whether two firms have prior links,
taking the value of 1 if there was previous transaction and 0 otherwise. The choice of this
variable concerns the length of time during which the previous transactional relationships
are likely to influence the current transactional links. One choice is to include all the
previous links in the transactional networks; while another possibility is to use “moving
window” (Gulati, 1995), that is to say; only transactional links established in certain
previous years will have influence on current transactional links. Considering the fact
that the repeated transactional relationships appeared periodically in my data and the
relative shortness of period observed (Gulati, 1995), I apply the first approach to include
any previous transactional link no matter when it occurred.
Previous transaction value
Previous transaction value may have influence on the existence and magnitude of
inter-firm transactions. The greater the previous transaction value between two firms, the
more likely that two firms will sustain their transactional relationship and increase future
transaction value. This control variable gauges the total value of previous transactions
between two firms.
The last concern relates to the treatment of the presence of few transactions that firms
have with themselves. The presence of same-firm transactions is contrary to my
expectation and may violate the regression estimates. Therefore, I include a dummy
variable to control for the effect.
52
Same-firm transaction
Though unexpected, some firms in my study sample have major transactions with
themselves, usually with a large amount of transaction value. This binary control variable
captures whether firms having transactions with themselves, taking the value of 1 if firms
have and 0 otherwise.
Table 4.4 Summary of Variables and Measures
Variables Measures
Dependent variables Y1: Existence of major transaction
Y2: Magnitude of major transaction
Independent variables X1:Dependency degree
X2: NAICS proximity
Control variables (Firm-level)
Control variable (Dyad-level)
C1: Number seller outgoing links
C2: Number buyer incoming links
C3: Value seller outgoing links
C4: Value buyer incoming links
C5:Seller revenues
C6:Buyer revenues
C7:Previous transaction between firms
C8:Previous transaction values between
firms
C9: Same-firm transaction
4.2.4 Descriptive Statistics and Correlation Matrix
Table 4.5 presents the descriptive statistics and correlation matrix for all the variables
involved in each of the two sets of regression. The data indicates the diversity of firms.
Especially for firm attributes, the ratios of “seller revenues”, “buyer revenues”, “outgoing
links (values)”, “incoming links (values)” suggest significant variance across firms
included in my sample, notwithstanding the fact that they are all large firms involved in
major transactions. The data on “previous transaction values” points to the significant
variance across different dyads. Generally, buyers are much larger than sellers regarding
the ratio of firm revenues.
The correlation matrix suggests that “seller revenues” is highly correlated with the
53
dependent variable “magnitude of major transactions” and with the dyad-level variable
“previous transaction values” between firms. And “buyer revenues” is moderately
correlated with the “value buyer incoming links”. Firms have transactions with larger
values in previous years are more likely to have transactions with larger values
subsequently. Despite high correlations in some cases, the correlations between
independent variables and control variables are generally low (below 0.3).
54
Table 4.5 Descriptive Statistics and Correlation Matrix
Variables Mean S.D. Min Max DV IV1 IV2 C1 C2 C3 C4 C5 C6
Model 1 (N=23586)
DV: Existence of transaction 0.03 0.15 0 1 1.00
IV1: Dependency -0.00 0.45 -1 1 0.17 1.00
IV2:NAICS Proximity 1.29 1.19 0 3 -0.16 -0.00 1.00
C1:Outgoing values 104.28 208.67 0 1725 0.05 0.09 0.06 1.00
C2:Incoming values 109.16 333.69 0 3008 0.34 0.25 -0.23 0.05 1.00
C3:Seller revenues 5117.55 13809 0.221 137634 -0.05 -0.24 -0.24 -0.06 -0.01 1.00
C4: Buyer revenues 5128.80 13813 0.221 137634 0.32 0.24 -0.24 0.01 0.56 -0.01 1.00
C5: Previous values 7.95 121.58 0 7588 0.34 0.07 -0.06 0.15 0.16 -0.00 0.09 1.00
C6: # previous transactions 0.02 0.15 0 1 0.75 0.16 -0.15 0.05 0.32 -0.04 0.33 0.43 1.00
Variables Mean S.D. Min Max DV IV1 IV2 C1 C2 C3 C4 C5 C6 C7
Model 2 (N=599)
DV: Value of transaction 112.82 211.45 0.02 1725.85 1.00
IV1: Dependency 0.47 0.19 -1 1 -0.04 1.00
IV2:NAICS Proximity 0.11 0.46 0 3 0.23 -0.32 1.00
C1:Outgoing links 1.78 0.88 1 5 -0.13 -0.19 -0.01 1.00
C2:Incoming links 10.17 7.61 1 27 -0.12 0.03 -0.26 -0.05 1.00
C3:Seller revenues 877.03 1843.62 0.22 19176 0.88 -0.04 0.02 -0.05 -0.13 1.00
C4: Buyer revenues 32981 31784 194.19 137634 -0.11 -0.01 -0.22 -0.02 0.71 -0.11 1.00
C5: Previous values 295.35 688.06 0 8065.65 0.79 0.02 0.12 -0.09 -0.16 0.79 -0.11 1.00
C6: # previous transactions 0.78 0.41 0 1 -0.08 0.02 -0.07 -0.04 0.06 0.06 0.08 0.23 1.00
C7:Same firm 0.01 0.09 0 1 0.45 -0.23 0.57 -0.08 -0.09 0.07 -0.09 0.25 0.00 1.00
55
4.3 Analytical Approach
In this section, I provide the rationale for my analytical approach and specify models
that used to test the hypotheses of the determinants of the presence and magnitude of
inter-firm major transactions.
4.3.1 Fixed Effect Models
Fixed-effect (FE) models completely ignore the between-group variation and focus
only on the within-group variation. The major advantage of FE-methods is the ability
to control for all stable characteristics of the subjects in the study, thereby eliminating
potential biases caused by time-constant unobserved heterogeneity problem (Baltagi,
2001; Wooldridge, 2002). However, FE-models have three main restrictions. First,
FE-regressions fail to estimate the effects of time-constant covariates. The
within-group transformation applies only with time-varying covariates (Wooldridge,
2002). Second, FE-models cannot estimate coefficients for variables that have no
within-subject variations. That is to say, there must be some variations in the
explanatory variables, otherwise; FE-regressions cannot estimate their effects. This is
a big problem if only a few observations show a change in an explanatory variable.
And practically, many software packages (e.g. STATA) automatically drop off the
explanatory variables with few variations in observations. Third, another disadvantage
of FE-regressions is that the number of unknown parameters increases with the
number of sample observations (Hsiao, 2007).
4.3.2 Random Effect Models
Random-effect (RE) models have become popular in a variety of disciplines ranging
from economic studies to social network studies. The main advantage of RE-estimator
is that it provides estimates for time-constant covariates. Since my study is interested
56
in reporting the effects of time-invariant variables, RE-methods are more appropriate
than FE-methods to analyze my data.
Hausman (1978) indicates that FE-estimator is usually consistent regardless of the
effects are random or fixed, while RE-estimator is only efficient and consistent when
random effects are not correlated with covariates. A key assumption of the
RE-estimator is that the random effects are uncorrelated with the observed
time-constant and time-varying covariates in the model (Hausman, 1978). Sometimes,
this assumption goes untested and may result in biased estimators. Hausman test
(Hausman, 1978) is used to differentiate between FE-models and RE-models in panel
data. The null hypothesis is that the coefficients estimated by the efficient
RE-estimator are the same as the ones estimated by the consistent FE-estimator, while
the alternative is that the RE-estimator is inconsistent and therefore the FE-estimator
is preferred.
Therefore, Hausman test was applied to determine which estimator is appropriate for
my panel data. For the panel data used to test the magnitude of major transactions,
Chi-square is 6.76 (p=0.56), therefore at 95% confidence level, I fail to reject the null
hypothesis. That is to say, RE-estimator is consistent in this case. Since RE- estimator
is more efficient than FE-estimator (Wooldridge, 2002), I use RE-estimator to predict
the magnitude of major transactions. For the panel data used to test the existence of
major transactions, Chi-square is 38.01 (p=0.00), therefore at 95% confidence level, I
reject the null hypothesis. That is to say, RE-estimator is inconsistent in this case.
Therefore, FE-estimator is applied to study the existence of major transactions.
4.3.3 Statistical Technique
For studying the existence of major transactions, I use the following fixed-effect
model for panel data:
57
= β + +
Where is the probability of a major transaction between firm i and firm j at time
t; β is the parameter vector; is the time-variant vector of covariates or
explanatory variables characterizing firms i and j, is the unknown intercept and
is the error term.
For studying the magnitude of major transactions, I used the following random-effect
model for panel data:
= β + + +
Where is the value of major transaction between firm i and firm j at time t; β is
the parameter vector; is the time-variant vector of covariates or explanatory
variables characterizing firm i and firm j, is the unknown intercept, is the
within-group error term, and is the between-group error term.
4.4 Descriptive Statistics
I present descriptive statistics in the following section. It provides a holistic view in
terms of the nature of my data including: 1) Trends for firms (sellers and buyers) from
manufacturing and service provider industries in apparel sector and from external
industries; 2) Number of unique NAICS codes (4-digit level) per five-year period for
sellers, buyers, and total firms; and 3) Number of transactions and value of
transactions for five-year period.
4.4.1 Data Records
The last column in this file tells the exact date when the data was inputted and
indicates that our data is tracked from 1976/11/30 to 2010/3/31. Thanks to the date
58
information, Figure 4.6 demonstrates the total observations for every five years. It
shows a clear rise from 1980 to 1995 and a peak from 1991 to 1995, but the following
period from 1995 to 2010 witnessed continuous drops every five years.
Figure 4.6 Number of All Firms and Firms from Manufacturing, Service, and External
Sectors Every Five Years
Figure 4.6 also reports the number of observations from manufacturing and service
sectors. Before 1995, the number of observations from manufacturing sector has
increased continuously. The number of observations has increased by approximately
100 for every five-year period from 1976 to 1995. However, after 1995, the number of
observations from manufacturing sector has dropped gradually. The trend of number
of observations from service provider sectors is consistent with the trend of
manufacturing sector. It is worth reporting that the trend line of number of
observations from service sector was catching up with the trend line in terms of
manufacturing sector at the period from 2006 to 2010. The number of observations
from external sectors has increased all along from 1976 to 2000 and then dropped
1976-
1980
1981-
1985
1986-
1990
1991-
1995
1996-
2000
2001-
2005
2006-
2010
Manufacturing sector 181 274 378 495 446 291 212
Service sector 144 232 356 415 338 238 207
External sector 39 72 96 106 130 109 93
All sectors 364 578 830 1016 914 638 512
181 274
378
495 446
291 212
144 232
356 415
338 238 207
39 72 96 106 130 109 93
364
578
830
1016
914
638
512
0
200
400
600
800
1000
1200
Manufacturing sector Service sector
External sector All sectors
59
slightly in the recent 10 years.
4.4.2 Number of Unique NAICS Codes (4-digit level) per Year
I identify number of 4-digit unique NIACS codes for sellers, buyers and total firms
for every five-year in figure 4.7. The trend lines for sellers and buyers are overlapping,
thus indicating that the numbers of unique NAICS codes for both kinds of firms are
quite close. After 2000, the trend lines of both are nearly in parallel and the number of
unique NAICS codes for buyers are always higher than the number for sellers. The
trend line for total firms is much above and number of unique NAICS for each year is
higher than numbers of unique NAICS for sellers and buyers, suggesting that sellers
and buyers appearing for every five years are from different industries at 4-digit
NAICS codes level. A limitation is that the SIC/NAICS system has evolved over this
time period.
Figure 4.7 Trend per Five Years for Number of Unique NAICS Codes (4-digit level)
of Sellers, Buyers and Total Firms
4.4.3 Number of Transactions and Value of Transactions
Among the identified 2431 transaction observations, there are 2030 observations
39
57
71
87 79
52
38
37
65
83 85 91
67 54
68
112
132
169
152
107
78
0
20
40
60
80
100
120
140
160
180
Seller NAICS
codes
Buyer NAICS
codes
Total firms
NAICS codes
60
having the exact amount of transaction value. Therefore, I only include the 2030
transaction observations here to depict the trend for value of transactions in a
five-year period of time in Figure 4.8. From 1976 to 1995, the trend for total value of
transactions is continuously increasing. After 2003, there is a five-year rising trend for
total value of transactions from 2006 to 2010. The year 1996 has the highest number
of transactions with a number of 97 and the year 2008 has the largest amount of
transaction value with a total value of 9567.212. It is worth noting that the value of
transactions has not decreased that much even though the number of transactions
dropped in the decade from 2000 to 2010.
Figure 4.8 Trends for Number of Transactions and Value of Transactions
157
239
314
418 402
278 217
4974.926 7839.363
11704.976
35880.769 30407.398 28111.748
32946.612
0
10000
20000
30000
40000
0
100
200
300
400
500
1976-1980 1981-1985 1986-1990 1991-1995 1996-2000 2001-2005 2006-2010
Number of transactions Value of transactions
61
5. Results
In this section, I first provide the empirical results from the analysis in terms of the
determinants of major inter-firm transactions. Then, a comparison of results from
different models is discussed. Finally, I provide the descriptive statistics regarding the
evolution of the U.S. apparel sector over the course of 30 years.
5.1 Determinants of the Presence of Inter-firm Transactions
The fixed-effect models are run by STATA 10.0 and the estimation results are reported
in Table 5.1. In Table 5.1, the dependent variable is a dummy that indicates the
existence of a major inter-firm transaction. Model 1 reports the effects of firm-level
and dyad-level control variables on the existence of major transactions. Models 2-4
present the results of the sequential introduction of independent variables of interest.
Model 5 is the full model including all independent variables and control variables.
Specifically, Model 3 tests whether there is an inverted-U shape relationship between
inter-firm similarity and the existence of major transaction.
According to Table 5.1, Hypothesis 1 is not supported. In Models 2, 3, and 5,
inter-firm similarity, as measured by NAICS proximity, is statistically significant but
is negatively associated with the existence of major transaction. Model 3 tests the
quadratic relationship between inter-firm similarity and the existence of inter-firm
transaction by introducing the squared term of NAICS proximity. It is somewhat
surprising that the results indicate that there is a U-shape relationship, rather than an
inverted-U shape relationship as hypothesized. Contrary to my expectation, the
empirical results suggest that firms that undertake more similar activities will be less
likely to have major transactions with each other. But firms within the same industry
may have transactional relations, which is inconsistent with the expectation that firms
in the same industry will compete – rather than cooperate – with each other.
62
Such inconsistency with Hypothesis 1 may be explained by the existence of firms that
develop permeable boundaries that allow these firms to selectively participate at
multiple points along the value chains (Jacobides & Billinger, 2006), while engaging
in tapered integration (Harrigan, 1985). For example, two apparel manufacturers with
permeable vertical architecture can concurrently buy, sell, and ally with each other.
Moreover, an alternative explanation can be the limitation of the measure (NAICS
proximity) that SIC-based measurement does not consistently reflect relationships
among valuable resources in the ways that firms actually combine them to create
value (Bryce & Winter, 2009).
Hypothesis 2 is supported by the empirical results set out in Table 5.1. Inter-firm
complementarity, measured as firms’ sector roles dependency, is positively associated
with the existence of major inter-firm transaction. Furthermore, this independent
variable is always statistically significant at the 99% confidence level. The results are
consistent with the expectation that, from a demand perspective, there exist
dependency relations between firms’ sector roles that are defined on the basis of
complementarity in activities. The empirical results indicate that, within the vertical
structure of value chains, inter-firm relations will be hierarchically ordered with firms
providing services to final customers at the highest level, with firms manufacturing
products at the intermediate level, and with firms supplying raw materials at the
lowest level.
The dyad-level control variables generally behave as expected. The two dyad-level
variables that control for the impact of previous links between buyers and sellers are
both statistically significant. The binary variable “previous transaction” (an indicator
of whether there are previous major transactions between two firms) has a large
positive effect on the likelihood of subsequent transaction. Similarly, “previous
transaction value” is positively related to the existence of major transaction. The
larger the transaction value between two firms in the past, the more likely there will
be one or more subsequent transactions. This is not surprising because firms are
63
willing to transact with past customers due to familiarity and mutual trust. In addition,
I include these two variables to capture dyad-level unobserved heterogeneity.
Table 5.1 also reports the estimation results for firm-level control variables. Not
surprisingly, “value seller outgoing links” and “value buyer incoming links” are both
statistically significant and are positively associated with the existence of major
transaction. These two variables capture the effect that firms with a large value of
outgoing/incoming links are more likely to have major transactions with
customers/suppliers. I also include two variables to control for firm size, namely
“seller revenues” and “buyer revenues.” All models show that the “seller revenues”
variable is negatively related to the existence of major transaction. Moreover, the
“buyer revenues” variable is insignificant in Models 2, 3, and 5. The results of the
firm revenues variables are inconsistent with the expectation that firms with larger
revenues are more likely to have major transactions and are better able to attract
potential partners. One of the plausible explanations can be the nature and limitations
of my data. My data are comprised of major transactions which are restricted to
situations where the sales (transaction) value to a specific buyer exceeds 10% of the
seller’s total annual revenues. In other words, the larger the seller’s revenues, the
lower the probability that the value of a particular transaction will exceed 10% of the
seller’s total annual revenues; in this situation, the transaction would not be reported
in my data. Another explanation is that firm size fails to be an appropriate predictor in
explaining firms’ likelihood to have major transactions, especially when compared
with other dyad-level predictors.
In terms of the existence of major inter-firm transactions, Hypothesis 1 is not
supported. Contrary to this hypothesis, there is a U-shape relationship between
inter-firm similarity and inter-firm transaction. Hypothesis 2 is fully supported by the
empirical results. It is worth mentioning that the control variables “previous
transaction” and “value buyer incoming links” are the strongest predictors, in
comparison with all variables.
64
5.2 Determinants of the Magnitude of Inter-firm Transactions
The random-effect models are run by STATA 10.0 and the estimation results are
presented in Table 5.2. The random-effect models generate a coefficient rho, which is
an indicator of the absence or presence of heterogeneity. An estimate close to zero
implies little heterogeneity, and at all times dependence in the occurrence of major
transactions can be ascribed to the independent variables in the models (Gulati, 1995).
In my regression, the parameter rho is close to zero, thus indicating the absence of the
heterogeneity problem in my panel models. Model 1 reports the effects of firm- and
dyad-level control variables on the magnitude of major transaction. Models 2-4
present the results of the sequential introduction of independent variables. Model 5 is
the full model including all independent and control variables. Specifically, Model 3
tests the quadratic relationship between inter-firm similarity and the magnitude of
major transaction.
Hypothesis 1 states that there will be an inverted-U relationship between inter-firm
similarity and inter-firm transaction. In Model 3, the “NAICS proximity” is positive
and significant and the squared term of NAICS proximity is negative and significant,
thus indicating the presence of an inverted-U relationship between NAICS proximity
and the magnitude of major transaction. In Model 2, inter-firm similarity, measured as
NAICS proximity, is not significant, but it becomes important in the later models
when more variables are included, thus suggesting a mediation effect as a result of the
addition of new variables. In Model 5, the squared term of “NAICS proximity” is not
significant when taking into account all the variables. There is a mixed result for
Hypothesis 1 that firms will be more likely to have transactions of larger value with
similar firms, rather than with unrelated firms or with equivalent peers from the same
industry.
Hypothesis 2 is fully supported by the empirical results shown in Table 5.2. The
65
coefficient of the dependency degree variable is significant and positive at the 99%
confidence level, thus indicating that inter-firm complementarity is positively
associated with the magnitude of major inter-firm transaction. In addition, the
introduction of the variable “dependency degree” increases the R^2 and the Wald
Chi^2 of the model. A comparison among models indicates that “dependency degree”
behaves better than “NAICS proximity” in predicting the dependent variable.
There are also interesting results for the control variables. First, the “previous
transaction” dummy (which is the indicator of whether there has been a prior
transaction) is insignificant across all models. The result suggests that whether there
has been a previous inter-firm transaction will not affect the magnitude of a
subsequent transaction. Such a result would be inconsistent with the expectation that a
previous transactional relationship would influence the current transaction because
firms tend to sustain the relationship and to increase future transaction value with
their old partners. The reason might be that such information may not be captured by
the major inter-firm transaction. Furthermore, in order to measure the dummy
“previous transaction,” I include all previous transactional links between firms
regardless of when they took place. However, it may be that only transactional links
established in certain prior years have an influence on the magnitude of the current
transaction. “Previous transaction value” between two firms is positively associated
with the magnitude of a major transaction. This is not surprising because firms that
had large-value transactions in previous years will subsequently sustain or strengthen
their relationship. It is worth noting that the “previous transaction value” variable
becomes less significant in later models when more variables are introduced, thus
indicating a suppresser effect by the addition of other variables. Second, the
coefficient of “number seller outgoing links” is significant but negative. It is
understandable that the size of a single transactional link will be reduced when the
number of transactional links increases, especially in terms of major transactions. As
expected, “seller revenue” is positively associated with the magnitude of major
inter-firm transaction. However, the variables “buyer revenue” and “number buyer
66
incoming links” are both insignificant across models in my regression. The results
may suggest that the buyer’s attributes are not important in explaining the magnitude
of major transactions.
In terms of the magnitude of major inter-firm transaction, Hypothesis 1 and
Hypothesis 2 are both supported. Both inter-firm similarity (as measured by NAICS
proximity) and inter-firm complementarity (as measured by the degree of dependency)
are significant in explaining major inter-firm relations. Table 5.2 suggests that the
magnitude of inter-firm transactions seems to be independent of the buyer’s attributes.
The seller’s attributes are more important than the buyer’s attributes in explaining the
magnitude of major inter-firm transactions.
5.3 Comparison of Results
The empirical results of the two sets of regressions are inconsistent. Hypothesis 1 is
supported when predicting the magnitude of inter-firm transactions, but is not
supported when predicting the existence of inter-firm transactions. Surprisingly, when
predicting the existence of transaction, the expected inverted-U shape relationship is
replaced by a U-shape relationship between inter-firm similarity and inter-firm
transaction.
The inconsistency with Hypothesis 1 can be ascribed to the limitation of the
SIC-based measure to effectively capture the nature of inter-firm similarity. SIC-based
measurement has been criticized because it fails to consistently reflect relationships
among valuable resources in the ways that firms actually combine them to create
value (Bryce & Winter, 2009). In my data, the preponderance of major transactions in
apparel sector is from manufacturer to retailer. Firms from NAICS subsector 315
(Apparel Manufacturing) and firms from NAICS subsector 448 (Clothing and
Clothing Accessories Stores) are closely related, but the degree of NAICS proximity
67
between firms from those two subsectors takes the value of 0. Therefore, for most of
the observations that take the value of 1 (which indicates the existence of a transaction)
for their binary dependent variable “existence of transaction,” their independent
variable “NAICS proximity” takes the value of 0. When predicting the existence of a
transaction, this problem will result in a negative coefficient of NAICS proximity
because most transactions in the apparel sector take place between firms that are in
fact related but whose connection is obscured by their positions in unrelated NAICS
industries. This unavoidably weakens the sensitivity of my data in terms of reflecting
the actual relation between inter-firm similarity and the existence of transaction. The
same problem occurs again when predicting the “magnitude of transaction.” But this
problem is mitigated because the “magnitude of transaction” is a continuous variable.
Another explanation for the unexpected U-shape relationship is the presence of
partially integrated firms. Firms develop a permeable vertical architecture that is
partly integrated and partly open to the markets along a firm’s value chain, thereby
allowing their units to make and buy simultaneously, as well as to transfer
downstream or sell (Jacobides & Billinger, 2006). Due to the benefits of vertical
permeability, firms within the same industry will have transactions with their peers in
order to effectively monitor competitors and competitively benchmark.
Hypothesis 2 is fully supported by the empirical results of both sets of regressions.
Inter-firm complementarity is always statistically significant and is positively
associated with both the existence and magnitude of major inter-firm transaction.
Overall, the results imply that inter-firm complementarity is a strong predicator in
explaining inter-firm relations and that it behaves better than inter-firm similarity in
explaining major inter-firm transactions.
There are some notable inconsistencies between the two sets of regressions regarding
the results of the control variables. First, “seller revenues” is negatively associated
with the existence of transaction, but is positively associated with the magnitude of
68
transaction. This difference may be attributed to the nature of my data. The criterion
of a major transaction is that the transaction value to a specific buyer must exceed 10%
of the seller’s total annual revenues. Therefore, the larger the seller’s revenues, the
less likely it is that a single transaction will be reported as a major transaction. Second,
the results do not show a significant relationship between previous transaction and the
magnitude of current transaction. While the existence of previous transactions will
induce subsequent transactions, it does not have a significant impact on the magnitude
of subsequent transactions. Third, while the buyer’s attributes have a positive
relationship with the existence of transaction, these attributes do not have a
statistically significant impact on the magnitude of transaction. The results imply that
the seller’s and buyer’s attributes are both important in explaining the likelihood of a
transaction. Yet the magnitude of transaction seems to be related only with the seller’s
attributes while being independent of the buyer’s attributes.
Given the joint consideration of inter-firm similarity and inter-firm complementarity,
the empirical results show that the existence and magnitude of major transaction will
be a reflection of inter-firm complementarity, rather than of inter-firm similarity.
Major transactions are more likely to take place between vertically related pairs of
industries to which firms belong than between horizontally equivalent peers in similar
industries. In addition, my results imply that the theoretical sector role dependency
concept will better predict inter-firm relations than the empirical relatedness measure
based on NAICS.
69
Table 5.1 Logistic Regression of the Existence of Transaction on NAICS Proximity and Dependency (N=23586)
Independent variables Model 1 Model 2 Model 3 Model 4 Model 5
Seller revenues -0.0004***
(-5.14)
-0.0004***
(-6.10)
-0.0004***
(-6.11)
-0.0004***
(-4.24)
-0.0004***
(-4.95)
Buyer revenues 0.00001**
(3.36)
0.000002
(0.78)
0.000001
(0.43)
0.00001**
(3.08)
0.000001
(0.52)
Outgoing values 0.002***
(4.33)
0.002***
(5.92)
0.002***
(5.85)
0.002***
(3.98)
0.002***
(5.20)
Incoming values 0.002***
(13.52)
0.001***
(10.36)
0.001***
(9.98)
0.002***
(12.13)
0.001***
(9.16)
Previous transaction 5.354***
(31.20)
4.799***
(27.53)
4.767***
(27.31)
5.113***
(29.78)
4.675***
(26.95)
Previous transaction values 0.0005*
(2.52)
0.0006**
(3.10)
0.0006**
(2.95)
0.0004*
(1.93)
0.0005*
(2.30)
NAICS proximity -1.247***
(-9.42)
-2.111***
(-6.60)
-2.113***
(-6.24)
NAICS proximity^2 0.375**
(3.13)
0.438**
(3.43))
Dependency degree 1.697***
(8.35)
1.336***
(4.92)
Log likelihood -1023.569 -954.492 -949.668 -986.459 -934.506
LR Chi^2 3362.38*** 3500.53*** 3510.18*** 3436.60*** 3540.50***
Note: t-value or chi-square value in parentheses; *: p<0.05; **: p<0.01; ***: p<0.001.
70
Table 5.2 GLS Regression of the Magnitude of Transaction on NAICS Proximity and Dependency (N=599)
Independent variables Model 1 Model 2 Model 3 Model 4 Model 5
Seller revenues 0.091***
(41.79)
0.091***
(41.67)
0.091***
(41.85)
0.093***
(43.63)
0.094***
(43.84)
Buyer revenues -0.000003
(-0.03)
0.00001
(0.07)
0.00001
(0.10)
0.00005
(0.46)
0.00006
(0.59)
Outgoing links -12.395***
(-4.58)
-12.602***
(-4.67)
-12.283***
(-4.56)
-8.992**
(-3.38)
-8.713***
(-3.28)
Incoming links 0.696
(1.47)
0.738
(1.64)
0.826
(1.83)
0.576
(1.34)
0.769
(1.77)
Previous transaction 7.584
(1.25)
7.328
(1.21)
7.234
(1.20)
8.446
(1.45)
9.294
(1.59)
Previous transaction values 0.022***
(3.55)
0.023***
(3.67)
0.022***
(3.59)
0.016**
(2.60)
0.015*
(2.47)
Same firm 854.842***
(31.18)
844.744***
(25.70)
876.655***
(24.35)
905.264***
(32.65)
890.186***
(25.57)
NAICS proximity 2.735
(0.43)
33.932*
(2.13)
32.510*
(2.21)
NAICS proximity^2
-13.673*
(-2.14)
-8.408
(-1.35)
Dependency degree 83.681***
(6.51)
88.421***
(6.63)
Constant 28.021**
(3.38)
27.737**
(3.41)
25.459**
(3.11)
-18.979
(-1.79)
-26.672*
(-2.39)
rho 0.021 0.00 0.00 0.00 0.00
R^2 0.927 0.927 0.928 0.932 0.933
Wald Chi^2 7331.32*** 7541.31*** 7591.74*** 8122.89*** 8189.97***
Note: t-value or chi-square value in parentheses; *: p<0.05; **: p<0.01; ***: p<0.001.
71
5.4 Evolution of Apparel Sector
Figure 5.1 presents the proportion of service-provider firms versus manufacturing
firms involved in major transactions in the apparel sector. According to Hypothesis 3,
over time there will be a shrinking proportion of manufacturer firms relative to
service providers involved in major transactions in the U.S. apparel sector. In the
apparel sector, value chains are dominated by retailers and marketers (Gereffi et al.,
2002; Gereffi & Frederick, 2010) that exert considerable power over wholesalers,
manufacturers, and textile providers. The competitive advantage of U.S. apparel firms
may lie in their ability to capitalize on high quality and maintain strong brand
identities (Mittelhauser, 1997), to exert substantial control over their resources and
profit distribution (Gereffi & Frederick, 2010), and to control high value-added
activities such as branding and retailing (Gereffi, 2002). Therefore, I expect to see an
increasing proportion of service activities versus manufacturing activities in the U.S.
apparel sector. This expectation is consistent with the view that the U.S. outsources
and offshores its manufacturing activities to less-developed areas in order to achieve
economic superiority by capturing the most downstream value-added (Tassey, 2010).
In Figure 5.4, from 1976 to 1990, the proportion of service firms versus
manufacturing firms has increased in every five-year period. On the other hand,
during the decade from 1991 to 2000, this proportion has decreased by 10% in each of
the component five-year periods. The decade from 2000 to 2010 witnessed a
substantial increase in service activities. Therefore, Hypothesis 3 is not fully
supported due to the mixed results. But, interestingly, my results indicate a wave
pattern in the trend associated with the proportion of service activities versus
manufacturing services over time.
72
Figure 5.1 Proportion of Service Firms versus Manufacturing Firms
Figure 5.2 presents the proportion of external firms involved in transactions with
apparel firms versus all firms. Besides some low fluctuations, the percentage of
external firms versus all firms is between 11% and 12% from 1976 to 1995. In the
subsequent period from 1996 to 2010, the proportion of external firms versus all firms
has increased steadily and it peaked during the 2006-2010 period. Hypothesis 4 is
supported by the fact that over time the proportion of major transactions that involve
firms outside the apparel sector tends to increase. This is not surprising in that there is
a continuous increase in the diversification of multiple players from non-apparel
industries because products and services currently are provided by networks of firms.
Empirical results have demonstrated that apparel suppliers increasingly invest in new
information technologies, distribution systems, and other associated services in order
to respond rapidly to consumer demand while minimizing exposure to inventory risk
(Abernathy et al., 1995).
79.56% 84.67%
94.18%
83.84% 75.78%
81.79%
97.64%
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
1976-1980 1981-1985 1986-1990 1991-1995 1996-2000 2001-2005 2006-2010
Proportion of service firms versus manufacturing firms
73
Figure 5.2 Proportion of External Firms versus All Firms
Figure 5.3 shows the proportion of non-regular transactions versus regular
transactions. Regarding regular transactions, I include one-level up transactions and
transactions between apparel manufacturers and retailers.3 In terms of non-regular
transactions, I include level-skipping, same-level, and backward transactions. My
result shows that the period from 1986 to 1990 witnessed the largest percentage of
non-regular transactions relative to regular transactions. Hypothesis 5 is not fully
supported because the trend appears to be random. But my result indicates the
existence of vertically integrated or tapered integrated firms, and the fact that firms
develop permeable boundaries. This finding is consistent with previous studies that
investigate “taper integration” (Harrigan, 1984), “partically integrated firms”
(Rothaermel et al., 2006), “permeable boundaries” (Jacobides & Billinger, 2006), and
“concurrent sourcing” (Parmigiani, 2007; Parmigiani & Mitchell, 2009).
3 According to apparel sector roles’ dependency, transactions between apparel manufacturers and retailers are
one-level skipping transactions. However, in the apparel sector, such transactions should be considered regular.
10.71%
12.46% 11.57%
10.43%
14.22%
17.08% 18.16%
0.00%
2.00%
4.00%
6.00%
8.00%
10.00%
12.00%
14.00%
16.00%
18.00%
20.00%
1976-1980 1981-1985 1986-1990 1991-1995 1996-2000 2001-2005 2006-2010
Proportion of external firms versus all firms
74
Figure 5.3 Proportion of Non-regular Transactions versus Regular Transactions
7.75% 9.25%
14.50%
10.39% 9.79%
12.86% 12.88%
0.00%
5.00%
10.00%
15.00%
20.00%
1976-1980 1981-1985 1986-1990 1991-1995 1996-2000 2001-2005 2006-2010
Proportion of non-regular transactions versus regular transactions
75
6. Conclusion
This study investigates the nature of major transactions in the U.S. apparel sector,
using a general theory that is applicable across sectors. The firm-level analysis
provides knowledge regarding the relationship between a firm’s value chain position
and the nature of its transactions with other firms. The industry-level analysis
examines the evolution, over the course of 30 years, of the apparel sector by focusing
on the relative importance of manufacturing activities versus service activities, and on
the involvement of firms from external sectors (non-apparel sectors). Business
practitioners and journalists have provided descriptive statistics and general
knowledge regarding the nature of inter-firm relations, but these are not theory-based
and so there is little possibility of generalization to other contexts. My study is able to
predict the existence and magnitude of major transactions between firms based on
theory.
Using a unique sample of over 2,000 observations of major inter-firm transactions, I
test the theory of inter-firm similarity and of inter-firm complementarity to explain
inter-firm relations. My results suggest that inter-firm relations are a consequence of
inter-firm complementarity, rather than of inter-firm similarities. Major transactions
are more likely to take place between the vertically related pairs of industries to which
firms belong, rather than between horizontally equivalent peers in similar industries.
In the discussion below, I present the contributions and implications of this study. The
last subsection discusses limitations and future research directions.
6.1 Contributions
This study provides a systematic description of the nature of major transactions in the
U.S. apparel sector, using a theory that has the potential to apply across different
sectors. This is one of the few studies to use an important dataset to examine the
76
nature and determinants of the overall network structure and the first to consider
network relations in addition to bilateral relations. This exceptional dataset has been
used by financial researchers to support their award-winning research, but it has never
been used by strategy researchers. The dataset holds great promise for examining the
structure and evolution of transactional networks across sectors over time.
In addition, this study examines the importance of the theory of inter-firm
complementarity, rather than the theory of inter-firm similarity, in explaining sector
architecture; therefore, it contributes to the literature by illuminating the positive and
statistically significant relationship between inter-firm complementarity and the
existence and magnitude of inter-firm transactions. Inter-industry similarity is
important for explaining the performance differences among firms in terms of
profitability (Rumelt, 1982), for determining the direction of firm growth and the
degree of diversification (Rumelt, 1974; Teece et al., 1994; Bryce & Winter, 2008),
and for predicting firms’ decisions to enter new markets (Lien & Klein, 2009).
Nevertheless, inter-firm similarity is not useful for predicting firms’ decisions to
establish transactional relations with other firms. My theory of inter-firm
complementarity is based on the theory presented in Dalziel’s (2007) paper, which
leads to the identification of complementary firm roles and which better predicts the
transactional relationships between firms that are critical to illuminating sector
architecture. Dalziel (2007) proposes a system-based approach to industry
classification and introduces the sector-role concept which distinguishes between
service and manufacturing subsectors, and between central and complementary firms.
My thesis is the first empirical test of the proposed theory that sectors are
hierarchically structured according to dependency relations.
Finally, my study elucidates the importance of complementarity and the limitation of
similarity in predicting inter-firm relations and hence transactional networks’ structure;
it therefore contributes to the future improvement of industry classification systems.
The current industry classification systems are built on the use of similarity in
77
activities as the criterion for aggregating lower-order groups into higher-order groups.
Such systems depict sector structures as homogeneous clusters of industries but, in
practical terms, firms from complementary industries are hierarchically ordered
according to dependency relations. My findings are consistent with previous studies
that have observed the failure of current industry classification systems to reflect the
structure of the economy (Bryce & Winter, 2009; Hicks, 2011) and the failure to
capture the broad patterns in the nature of inter-firm relations that have been indicated
by researchers investigating supply chains, ecosystems, and modular clusters
(Baldwin & Clark, 2000).
6.2 Implications
Transactions are the most basic economic activities of firms and have extensive
implications on firms’ social activities and public actions in terms of international
norms, ethical standards, sustainable development, and so forth. For instance, a
buy-sell transaction between a Western retailer and manufacturing factories in
less-developed countries has implications for employee welfare, the enforcement of
building standards, and labour safety issues in poverty-stricken areas. In 2010,
Foxconn, a Taiwanese company, was plagued by a series of suicides that resulted in
14 deaths. Such out-of-the-ordinary fatalities threw an awkward spotlight on the
labour practices of Foxconn, which is a unit of Taiwan’s Hon Hai Precision Industry
Co. Ltd., whose clients include Apple Inc., Hewlett-Packard Co., and Sony Ericsson
(Reuters, 2010). In addition to the series of tragedies that occurred at Foxconn’s plants,
a series of tragedies took place in the buyer firms’ supply chains. As Foxconn’s
overseas partners, Apple, Dell, and HP were indirectly involved in the fatalities and
made statements that they would further investigate and audit the working conditions
of their Asian suppliers (Dean, 2010). Fatalities of this nature are not restricted to the
high-technology sector. Recently, public media has reported the collapse of
garment-manufacturing buildings in Bangladesh and the deaths of at least 1,100
78
workers (Nolen, 2013). Western retailers and consumers who bought shirts made in
Bangladesh cannot isolate themselves from this tragedy since they are aware that their
purchasing activities and their pursuit of fast fashion and low cost to some extent
contributed to the fatalities in that country. However, this awareness is not based on
any official data pertaining to inter-firm transactions, but is instead based on reports
disseminated by the public media. Regrettably, it is only when tragedies and fatalities
take place that most people are informed of the fact that their purchasing of cheap
clothes exacts a heavy human cost.
In addition to the implications for human welfare, inter-firm transactions may also
provide information on environmental welfare and sustainable development. In order
to fully account for the environmental effects of the products and services that we
consume, we require readily accessible data on the relationships between firms in
global supply chains (Henrickson et al., 2006). In global value chains, most
developing countries concentrate their activities in lower value-added sectors such as
manufacturing and raw material industries. Developing countries have long been
blamed for their contributions to industrial pollution, especially to the rise of carbon
dioxide emissions. However, studies have suggested that the stabilization of emissions
in developed countries was partially due to the growth of imports from developing
countries (Peters et al., 2011) and have revealed the full extent of the West’s
responsibility for expanding the world’s emissions of greenhouse gases (Clark, 2009).
According to a report from Oslo’s Centre for International Climate and Environmental
Research, around 9% of total Chinese emissions are now the result of manufacturing
goods for the U.S., and around 6% are the result of producing goods for Europe.
Academics and campaigners increasingly say that responsibility for this pollution lies
with the consumer countries (Clark, 2009). Peters et al. (2011) have examined the
growth in emissions transfer via international trade from developing to developed
countries and their results indicate that international trade is a significant factor in
explaining the shift in emissions from both a production and a consumption
79
perspective (Peters et al., 2011). The failure of the Copenhagen summit may partially
be explained by the fact that the developing world is reluctant to take the West’s share
of the responsibility for causing emissions by means of off-shored manufacturing
facilities aimed at producing goods for foreign markets (Clark, 2009). To solve the
issue of off-shored emissions, a new challenge is introduced because of the
measurement of domestic, exported, and imported emissions from both the production
and consumption perspectives. Relevant international-trade data of each country is
required for such meticulous measurements. Data regarding inter-firm transactions are
of benefit to the calculation of emissions from both the production and consumption
perspectives and may provide empirical evidence that can be used for the further
improvement of environmental and climate-change policies. My study calls for efforts
to improve the current industry classification systems. If industry classification
systems were designed in ways that reflect actual industry structures, as manifested by
transactions, then such systems would be an effective means of providing a superior
knowledge of industry and international-trade policies, of human-welfare issues in
manufacturing industries, and of environmental issues in raw-material industries.
Second, the findings of this study also have several important implications regarding
the evolution of U.S. apparel sector and its future governance. Some analysts have
suggested a bleak future for the U.S. apparel sector based on the erosion of domestic
employment and the increase in import penetration (Mittelhauser, 1997). It has often
been reported that the availability of high-quality jobs has been compromised by the
outsourcing of manufacturing activities (Tassey, 2010). In order to better understand
the relationship between outsourcing and employment opportunities, we also need to
consider the growth of related professional service activities. The results of my study
indicate an increasing proportion of service activities versus manufacturing activities
in the apparel sector and an increasing proportion of external firms involved in
transactions with firms in the apparel sector. The methodology employed in my study
raises the possibility that the loss of manufacturing jobs has been compensated by a
gain in service jobs. In the apparel sector, value chains are dominated by retailers and
80
marketers that exert more power over wholesalers, manufacturers, and textile
providers. The competitive advantage of U.S. apparel firms may lie in their ability to
capitalize on high quality and maintain strong brand names (Mittelhauser, 1997), to
develop information integrated supply chains (Abernathy et al., 1995, 2000), to exert
substantial control over their resources and profit distribution (Gereffi & Frederick,
2010), and to control high value-added activities such as branding and retailing (Bair
& Gereffi, 2003). My study indicates a more encouraging picture of the U.S. apparel
sector than the pessimistic prognosis suggested by previous studies.
Finally, the durability of major transactions may have implications for firm
competition policies. The enduring linkages between certain large firms may indicate
the potential risk of dominance and monopoly in the market and may predict future
merger and acquisition activities.
6.3 Limitations and Future Research Directions
Despite this study’s attempt to be as thorough and rigorous as possible, there are
limitations that need to be acknowledged. First, due to the time limitation, this study
only investigates one sector. Therefore, the generalization of the findings in terms of
sector architecture and evolution in the apparel sector may not apply to other sectors.
My study is at the starting point of investigation into sector and value chain
architectures in the economy; future studies are expected to test the theory in more
complex sectors than the apparel sector. Studies on other sectors may display more
complex sector architecture, different transactional relationships, or tremendous
changes over time.
Second, the data sample of this study only contains “major transactions” rather than
the “total transactions” in the U.S. apparel sector. I believe that major transactions are
sufficient to capture the general picture of sector architecture and that large firms in
81
most cases lead the trend while small firms are more likely to follow. For example,
Luo et al. (2012) investigate the transaction networks of two large sectors in Japan
and demonstrate that there are no backward linkages or cycles once the 10 largest
firms are removed from the dataset in the Japanese electronics sector. However, a
further exploration with additional data on “total transactions” will yield a more
holistic view of the structure of transaction networks. Future studies are expected to
compare the differences or similarities between the network structure of major
transactions and the network structure of minor transactions.
Moreover, this study only considers one highly developed country in North America.
Inter-firm transaction networks in the apparel sector can be very different in Asia and
South America where most firms are positioned in low value-added industries and are
engaged in manufacturing activities. Future studies are expected to explore the
apparel sector’s architecture and transactional networks in the developing world.
Finally, in terms of the theory employed in my study, I believe that it will hold for
other sectors as well. Notwithstanding the strategic behaviour of firms, differences in
business environments across regions and nations, and relentless change in technology,
there are broad and stable patterns in the nature of inter-firm relations that can be
described succinctly. Future studies are therefore expected to empirically test the same
theory in other sectors and in other countries.
82
References
Abernathy, F. H., Volpe, A., & Weil, D. (2006). The future of the apparel and textile
industries: Prospects and choices for public and private actors. Environment and
Planning A, 38(12), 2207-2232.
Abernathy, F. H., Dunlop, J. T., Hammond, J. H., & Weil, D. (2000). Retailing and
supply chains in the information age. Technology in Society, 22(1), 5-31.
Abernathy, F. H., Dunlop, J. T., Hammond, J. H., Weil, D., Bresnahan, T. F., &
Pashigian, B. P. (1995). The information-integrated channel: A study of the U.S.
apparel industry in transition. Brookings Papers on Economic Activity.
Microeconomics, 1995, 175-246.
Arora, A., & Gambardella, A. (1990). Complementarity and external linkages: The
strategies of the large firms in biotechnology. The Journal of Industrial Economics,
38(4), 361-379.
Bain, J. S. (1968). Industrial organization. New York: Wiley.
Bair, J., & Gereffi, G. (2003). Upgrading, uneven development, and jobs in the North
American apparel industry. Global Networks, 3(2), 143-169
Baldwin, C. Y. (2000). In Clark K. B. (Ed.), Design rules: The power of modularity.
Cambridge, Mass.: MIT Press.
Baldwin, C. Y. (2008). Where do transactions come from? Modularity, transactions,
and the boundaries of firms. Industrial and Corporate Change, 17(1), 155-195.
Baltagi, B. H. (2001). Econometric Analysis of Panel Data. 2d Ed. New York: John
83
Wiley & Sons.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of
Management, 17(1), 99-120.
Bonaccorsi, A., & Giuri, P. (2001). The long-term evolution of vertically-related
industries. International Journal of Industrial Organization, 19(7), 1053-1083.
Bonaccorsi, A., & Giuri, P. (2001). Network structure and industrial dynamics. The
long term evolution of the aircraft-engine industry. Structural Change and Economic
Dynamics, 12(2), 201-233.
Borgatti, S.P., Everett, M.G. &Freeman, L.C. (2002). Ucinet for Windows: Software
for Social Network Analysis. Harvard, MA: Analytic Technologies
Brusoni, S., Prencipe, A., & Pavitt, K. (2001). Knowledge specialization,
organizational coupling, and the boundaries of the firm: Why do firms know more
than they make? Administrative Science Quarterly, 46(4), 597-621.
Bryce, D. J., & Winter, S. G. (2009). A general interindustry relatedness index.
Management Science, 55(9), 1570-1585.
Cacciatori, E., & Jacobides, M. G. (2005). The dynamic limits of specialization:
Vertical integration reconsidered. Organization Studies, 26(12), 1851-1883.
Clark. D. (2009). West blamed for rapid increase in China's CO2. The Guardian.
Available at:
http://www.guardian.co.uk/environment/2009/feb/23/china-co2-emissions-climate
Cohen, L., & Frazzini, A. (2008). Economic links and predictable returns. Journal of
84
Finance, 63(4), 1977-2011.
Dalziel, M. (2007). A systems-based approach to industry classification. Research
Policy, 36(10), 1559-1574.
Dalziel, M., & Zhang, S. C. (2008). The main and contingent effects of the systems
integrator and component supplier roles. Proceedings of the 4th IEEE International
Conference on Management of Innovation and Technology, ICMIT, 854-859.
Dean,J. (2010). Suicides Spark Inquiries: Apple, H-P to examine Asian supplier after
string of deaths at factory. The Wall Street Journal. Available at:
http://online.wsj.com/article/SB10001424052748704026204575267603576594936.ht
ml
DFAIT(Foreign Affairs and International Trade Canada) . (2011). See Sydor 2011
Dickerson, K. G. (1995). Textiles and apparel in the global economy, 2nd edition.
Englewood Cliffs, NJ: Prentice Hall.
Fixson, S. K., & Park, J. (2008). The power of integrality: Linkages between product
architecture, innovation, and industry structure. Research Policy, 37(8), 1296-1316
Frederick, S., & Gereffi, G. (2011). Upgrading and restructuring in the global apparel
value chain: Why China and Asia are outperforming Mexico and Central America.
International Journal of Technological Learning, Innovation and Development, 4(1-3),
67-95.
Gereffi, G. (1999). International trade and industrial upgrading in the apparel
commodity chain. Journal of International Economics, 48(1), 37-70.
85
Gereffi, G., Humphrey, J., & Sturgeon, T. (2005). The governance of global value
chains. Review of International Political Economy, 12(1), 78-104.
Gerefi, G., & Tam, T. (1998). Industrial upgrading through organizational chains:
dynamics of rent, learning-by-doing, and mobility in the global economy. 93rd Annual
Meeting of the American Sociological Association, San Francisco, California
Gulati, R. (1995). Social structure and alliance formation patterns: A longitudinal
analysis. Administrative Science Quarterly, 40(4), 619-652.
Gulbrandsen, B., Sandvik, K., & Haugland, S. A. (2009). Antecedents of vertical
integration: Transaction cost economics and resource-based explanations. Journal of
Purchasing and Supply Management, 15(2), 89-102.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica, 46(6),
1251-1271.
Hendrickson, C. T., Lave, L. B., & Matthews, H. S. (2006). Environmental Life Cycle
Assessment of Goods and Services: An Input-Output Approach. Resources for the
Future Press.
Hertzel, M. G., Li, Z., Officer, M. S., & Rodgers, K. J. (2008). Inter-firm linkages and
the wealth effects of financial distress along the supply chain. Journal of Financial
Economics, 87(2), 374-387.
Hicks, D. (2011). Structural change and industrial classification. Structural Change
and Economic Dynamics, 22(2), 93-105.
Hsiao, C. (2007). Panel data analysis—Advantages and challenges. Test, 16(1), 1-22.
86
Van Assche, A. with commentary by Todd Evans. (2012). Global value chains and
Canada’s trade policy: Business as usual or paradigm shift? IRPP study 32. Montreal:
Institute for Research on Public Policy
Jacobides, M. G. (2005). Industry change through vertical disintegration: How and
why markets emerged in mortgage banking. Academy of Management Journal, 48(3),
465-498.
Jacobides, M. G., & Billinger, S. (2006). Designing the boundaries of the firm: From
"make, buy, or ally" to the dynamic benefits of vertical architecture. Organization
Science, 17(2), 249-261.
Jacobides, M. G., Knudsen, T., & Augier, M. (2006). Benefiting from innovation:
Value creation, value appropriation and the role of industry architectures. Research
Policy, 35(8 SPEC. ISS.), 1200-1221.
Jacobides, M. G., & Winter, S. G. (2005). The co‐evolution of capabilities and
transaction costs: Explaining the institutional structure of production. Strategic
Management Journal, 26(5), 395-413.
Lamprinopoulou, C., & Tregear, A. (2011). Inter-firm relations in SME clusters and
the link to marketing performance. Journal of Business & Industrial Marketing, 26(6),
421-429.
Langlois, R. N., & Robertson, P. L. (1992). Networks and innovation in a modular
system: Lessons from the microcomputer and stereo component industries. Research
Policy, 21(4), 297-313.
Lien, L., & Klein, P. (2009). Using competition to measure relatedness. Journal of
Management, 35(4), 1078-1107.
87
Lin, Z. (., Yang, H., & Arya, B. (2009). Alliance partners and firm performance:
Resource complementarity and status association. Strategic Management Journal,
30(9), 921-940.
Luo, J.L., Baldwin, C.Y., Whitney, D.E., & Magee, C.L. (2012). The architecture of
transaction Networks: a comparative analysis of hierarchy in two sectors. Industrial
and Corporate Change, 21(6), 1307–1335.
Luo, J.L., & Magee, C. L. (2011). Detecting evolving patterns of self-organizing
networks by flow hierarchy measurement. Complexity, 16(6), 53-61.
MacCormack, A., Rusnak, J., & Baldwin, C. Y. (2006). Exploring the structure of
complex software designs: An empirical study of open source and proprietary code.
Management Science, 52(7), 1015-1030.
Mittelhauser, M. (1997) Employment Trends in Textiles and Apparel, 1973-2005.
Monthly Labor Review, August 1997, 24-35.
Neffke, F., & Henning, S. M. (2008). Revealed Relatedness: Mapping Industry
Space.Working Paper Series 08.19, Papers in Evolutionary Economic Geography,
Utrecht University, Utrecht, the Netherlands
Nelson, R. R. (1982). In Winter S. G. (Ed.), An evolutionary theory of economic
change. Cambridge, Mass: Belknap Press of Harvard University Press.
Nolen, S. (2013). Cheap clothes have helped fuel social revolution in Bangladesh. The
Globe and Mail. Available at:
http://www.theglobeandmail.com/news/world/cheap-clothes-have-helped-fuel-social-r
evolution-in-bangladesh/article11589450/
Parmigiani, A. (2007). Why do firms both make and buy? An investigation of
88
concurrent sourcing. Strategic Management Journal, 28(3), 285-311.
Parmigiani, A., & Mitchell, W. (2009). Complementarity, capabilities, and the
boundaries of the firm: The impact of within-firm and interfirm expertise on
concurrent sourcing of complementary components. Strategic Management Journal,
30(10), 1065-1091.
Penrose, E.T. (1959). The Theory of the Growth of the Firm. Blackwell: Oxford
Peters, G., Minx, J., Weber, C.L., & Edenhofer, O. (2011). Growth in emission
transfers via international trade from 1990 to 2008. Proceedings of the National
Academy of Sciences (PNAS) 108(21): 8903-8908
Podolny, J. M. (1994). Market uncertainty and the social character of economic
exchange. Administrative Science Quarterly, 39(3), 458-483.
Porter, M. E. (1985). Competitive advantage: Creating and sustaining superior
performance. London; New York: Collier Macmillan; Free Press.
Reuters. (2010).Foxconn worker plunges to death at China plant: report. Available at:
http://www.reuters.com/article/2010/11/05/us-china-foxconn-death-idUSTRE6A41M
920101105
Robins, J. A., & Wiersema, M. F. (2003). The measurement of corporate portfolio
strategy: Analysis of the content validity of related diversification indexes. Strategic
Management Journal, 24(1), 39-59
Rosenkopf, L., & Schilling, M. A. (2007). Comparing alliance network structure
across industries: Observations and explanations. Strategic Entrepreneurship Journal,
1(3-4), 191-209.
89
Rothaermel, F. T., Hitt, M. A., & Jobe, L. A. (2006). Balancing vertical integration
and strategic outsourcing: Effects on product portfolio, product success, and firm
performance. Strategic Management Journal, 27(11), 1033-1056.
Rothaermel, F. T. (2001). Complementary assets, strategic alliances, and the
incumbent’s advantage: An empirical study of industry and firm effects in the
biopharmaceutical industry. Research Policy, 30(8), 1235-1251.
Rumelt, R.P. (1974). Strategy, Structure and Economic Performance. Cambridge, MA:
Division of Research, Graduate School of Business Administration, Harvard
University.
Rumelt, R. P. (1982). Diversification strategy and profitability. Strategic Management
Journal, 3(4), 359-369.
Schilling, M. A., & Steensma, H. K. (2001). The use of modular organizational forms:
An industry-level analysis. The Academy of Management Journal, 44(6), 1149-1168.
Smith, A. (1817). In Garnier (Ed.), An inquiry into the nature and causes of the wealth
of nations [electronic resource] with a life of the author : Also, a view of the doctrine
of smith, compared with that of the French economists, with a method of facilitating
the study of his works, from the French of M garnier. Edinburgh: Printed for S. Doig
and A. Stirling.
Sorenson, O., & Stuart, T. E. (2001). Syndication networks and the spatial distribution
of venture capital investments. The American Journal of Sociology, 106(6),
1546-1588.
Strauss, M. (2013). Lower tariffs, fewer quotas make Bangladesh the go-to apparel
producer. The Globe and Mail. Available at:
90
http://www.theglobeandmail.com/news/world/lower-tariffs-fewer-quotas-make-bangla
desh-the-go-to-apparel-producer/article11589221/
Stuart, T. E. (1998). Network positions and propensities to collaborate: An
investigation of strategic alliance formation in a high-technology industry.
Administrative Science Quarterly, 43(3), 668-698.
Sturgeon, T., Van Biesebroeck, J., & Gereffi, G. (2008). Value chains, networks and
clusters: Reframing the global automotive industry. Journal of Economic Geography,
8(3), 297-321.
Sturgeon, T. J., & Gereffi, G. (2009). Measuring success in the global economy:
International trade, industrial upgrading, and business function outsourcing in global
value chains. Transnational Corporations, 18(2), 1-35.
Sturgeon, T. J. (2002). Modular production networks: A new american model of
industrial organization. Industrial and Corporate Change, 11(3), 451-496.
Tanriverdi, H., & Venkatraman, N. (2005). Knowledge relatedness and the
performance of multibusiness firms. Strategic Management Journal, 26(2), 97-119.
Tassey, G. (2010). Rationales and mechanisms for revitalizing US manufacturing
R&D strategies. Journal of Technology Transfer, 35(3), 283-333.
Teece, D. J. (2010). Forward integration and innovation: Transaction costs and
beyond. Journal of Retailing, 86(3), 277-283.
Teece, D. J., Rumelt, R., Dosi, G., & Winter, S. (1994). Understanding corporate
coherence. Theory and evidence. Journal of Economic Behavior and Organization,
23(1), 1-30.
91
Tomlinson, P. R. (2010). Co-operative ties and innovation: Some new evidence for
UK manufacturing. Research Policy, 39(6), 762-775.
Williamson, O. E. (1975). Markets and hierarchies, analysis and antitrust
implications: A study in the economics of internal organization. New York: Free
Press.
Williamson, O. E. (1985). The economic institutions of capitalism: Firms, markets,
relational contracting. London; New York: Collier Macmillan; Free Press.
Winter, S. G. (1988). On coase, competence, and the corporation. Journal of Law,
Economics, and Organization, 4(1), 163-180.
Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data.
Cambridge, MA: MIT Press.
Yamakawa, Y., Yang, H., & Lin, Z. (. (2011). Exploration versus exploitation in
alliance portfolio: Performance implications of organizational, strategic, and
environmental fit. Research Policy, 40(2), 287-296.
